IAIFI Papers

View high energy physics IAIFI papers on INSPIRE

Foundational AI Papers Theoretical Physics Papers Experimental Physics Papers Astrophysics Papers

Symbolic Regression with Multimodal Large Language Models and Kolmogorov Arnold Networks
Thomas R. Harvey, Fabian Ruehle, Cristofero S. Fraser-Taliente, James Halverson
[ arXiv:2505.07956 | code ]
Foundational AI

Abstract We present a novel approach to symbolic regression using vision-capable large language models (LLMs) and the ideas behind Google DeepMind's Funsearch. The LLM is given a plot of a univariate function and tasked with proposing an ansatz for that function. The free parameters of the ansatz are fitted using standard numerical optimisers, and a collection of such ansätze make up the population of a genetic algorithm. Unlike other symbolic regression techniques, our method does not require the specification of a set of functions to be used in regression, but with appropriate prompt engineering, we can arbitrarily condition the generative step. By using Kolmogorov Arnold Networks (KANs), we demonstrate that ``univariate is all you need'' for symbolic regression, and extend this method to multivariate functions by learning the univariate function on each edge of a trained KAN. The combined expression is then simplified by further processing with a language model.

I-Con: A Unifying Framework for Representation Learning
Shaden Alshammari, John Hershey, Axel Feldmann, William T. Freeman, Mark Hamilton
ICLR 2025 [ arXiv:2504.16929 | code ] published
Foundational AI

Abstract As the field of representation learning grows, there has been a proliferation of different loss functions to solve different classes of problems. We introduce a single information-theoretic equation that generalizes a large collection of modern loss functions in machine learning. In particular, we introduce a framework that shows that several broad classes of machine learning methods are precisely minimizing an integrated KL divergence between two conditional distributions: the supervisory and learned representations. This viewpoint exposes a hidden information geometry underlying clustering, spectral methods, dimensionality reduction, contrastive learning, and supervised learning. This framework enables the development of new loss functions by combining successful techniques from across the literature. We not only present a wide array of proofs, connecting over 23 different approaches, but we also leverage these theoretical results to create state-of-the-art unsupervised image classifiers that achieve a +8% improvement over the prior state-of-the-art on unsupervised classification on ImageNet-1K. We also demonstrate that I-Con can be used to derive principled debiasing methods which improve contrastive representation learners.

High-performance training and inference for deep equivariant interatomic potentials
Chuin Wei Tan, Marc L. Descoteaux, Mit Kotak, Gabriel de Miranda Nascimento, Seán R. Kavanagh, Laura Zichi, Menghang Wang, Aadit Saluja, Yizhong R. Hu, Tess Smidt, Anders Johansson, William C. Witt, Boris Kozinsky, Albert Musaelian
[ arXiv:2504.16068 | code ]
Foundational AI

Abstract Machine learning interatomic potentials, particularly those based on deep equivariant neural networks, have demonstrated state-of-the-art accuracy and computational efficiency in atomistic modeling tasks like molecular dynamics and high-throughput screening. The size of datasets and demands of downstream workflows are growing rapidly, making robust and scalable software essential. This work presents a major overhaul of the NequIP framework focusing on multi-node parallelism, computational performance, and extensibility. The redesigned framework supports distributed training on large datasets and removes barriers preventing full utilization of the PyTorch 2.0 compiler at train time. We demonstrate this acceleration in a case study by training Allegro models on the SPICE 2 dataset of organic molecular systems. For inference, we introduce the first end-to-end infrastructure that uses the PyTorch Ahead-of-Time Inductor compiler for machine learning interatomic potentials. Additionally, we implement a custom kernel for the Allegro model's most expensive operation, the tensor product. Together, these advancements speed up molecular dynamics calculations on system sizes of practical relevance by up to a factor of 18.

Learning Topological Invariance
James Halverson, Fabian Ruehle
[ arXiv:2504.12390 ]
Foundational AI

Abstract Two geometric spaces are in the same topological class if they are related by certain geometric deformations. We propose machine learning methods that automate learning of topological invariance and apply it in the context of knot theory, where two knots are equivalent if they are related by ambient space isotopy. Specifically, given only the knot and no information about its topological invariants, we employ contrastive and generative machine learning techniques to map different representatives of the same knot class to the same point in an embedding vector space. An auto-regressive decoder Transformer network can then generate new representatives from the same knot class. We also describe a student-teacher setup that we use to interpret which known knot invariants are learned by the neural networks to compute the embeddings, and observe a strong correlation with the Goeritz matrix in all setups that we tested. We also develop an approach to resolving the Jones Unknot Conjecture by exploring the vicinity of the embedding space of the Jones polynomial near the locus where the unknots cluster, which we use to generate braid words with simple Jones polynomials.

On Soft Clustering For Correlation Estimators: Model Uncertainty, Differentiability, and Surrogates
Edward Berman, Sneh Pandya, Jacqueline McCleary, Marko Shuntov, Caitlin Casey, Nicole Drakos, Andreas Faisst, Steven Gillman, Ghassem Gozaliasl, Natalie Hogg, Jeyhan Kartaltepe, Anton Koekemoer, Wilfried Mercier, Diana Scognamiglio, COSMOS-Web: The JWST Cosmic Origins Survey
[ arXiv:2504.06174 | code ]
Astrophysics

Abstract Properly estimating correlations between objects at different spatial scales necessitates (n2) distance calculations. For this reason, most widely adopted packages for estimating correlations use clustering algorithms to approximate local trends. However, methods for quantifying the error introduced by this clustering have been understudied. In response, we present an algorithm for estimating correlations that is probabilistic in the way that it clusters objects, enabling us to quantify the uncertainty caused by clustering simply through model inference. These soft clustering assignments enable correlation estimators that are theoretically differentiable with respect to their input catalogs. Thus, we also build a theoretical framework for differentiable correlation functions and describe their utility in comparison to existing surrogate models. Notably, we find that repeated normalization and distance function calls slow gradient calculations and that sparse Jacobians destabilize precision, pointing towards either approximate or surrogate methods as a necessary solution to exact gradients from correlation functions. To that end, we close with a discussion of surrogate models as proxies for correlation functions. We provide an example that demonstrates the efficacy of surrogate models to enable gradient-based optimization of astrophysical model parameters, successfully minimizing a correlation function output. Our numerical experiments cover science cases across cosmology, from point spread function (PSF) modeling efforts to gravitational simulations to galaxy intrinsic alignment (IA).

Quantum Mechanics and Neural Networks
Christian Ferko, James Halverson
[ arXiv:2504.05462 ]
Theoretical Physics

Abstract We demonstrate that any Euclidean-time quantum mechanical theory may be represented as a neural network, ensured by the Kosambi-Karhunen-Loève theorem, mean-square path continuity, and finite two-point functions. The additional constraint of reflection positivity, which is related to unitarity, may be achieved by a number of mechanisms, such as imposing neural network parameter space splitting or the Markov property. Non-differentiability of the networks is related to the appearance of non-trivial commutators. Neural networks acting on Markov processes are no longer Markov, but still reflection positive, which facilitates the definition of deep neural network quantum systems. We illustrate these principles in several examples using numerical implementations, recovering classic quantum mechanical results such as Heisenberg uncertainty, non-trivial commutators, and the spectrum.

Extracting the distribution amplitude of pseudoscalar mesons using the HOPE method
S.-P. Alex Chang, William Detmold, Anthony V. Grebe, Issaku Kanamori, C.-J. David Lin, Robert J. Perry, Yong Zhao
[ arXiv:2503.12198 ]
Theoretical Physics

Abstract The pseudoscalar meson light-cone distribution amplitudes (LCDAs) are essential non-perturbative inputs for a range of high-energy exclusive processes in quantum chromodynamics. In this proceedings, progress towards a determination of the low Mellin moments of the pion and kaon LCDAs by the HOPE Collaboration is reported.

Generative Modeling for Mathematical Discovery
Jordan S. Ellenberg, Cristofero S. Fraser-Taliente, Thomas R. Harvey, Karan Srivastava, Andrew V. Sutherland
[ arXiv:2503.11061 | code ]
Foundational AI

Abstract We present a new implementation of the LLM-driven genetic algorithm {funsearch}, whose aim is to generate examples of interest to mathematicians and which has already had some success in problems in extremal combinatorics. Our implementation is designed to be useful in practice for working mathematicians; it does not require expertise in machine learning or access to high-performance computing resources. Applying {funsearch} to a new problem involves modifying a small segment of Python code and selecting a large language model (LLM) from one of many third-party providers. We benchmarked our implementation on three different problems, obtaining metrics that may inform applications of {funsearch} to new problems. Our results demonstrate that {funsearch} successfully learns in a variety of combinatorial and number-theoretic settings, and in some contexts learns principles that generalize beyond the problem originally trained on.

L2M: Mutual Information Scaling Law for Long-Context Language Modeling
Zhuo Chen, Oriol Mayné i Comas, Zhuotao Jin, Di Luo, Marin Soljačić
[ arXiv:2503.04725 | code ]
Foundational AI

Abstract We rigorously establish a bipartite mutual information scaling law in natural language that governs long-range dependencies. This scaling law, which we show is distinct from and scales independently of the conventional two-point mutual information, is the key to understanding long-context language modeling. Using this scaling law, we formulate the Long-context Language Modeling (L2M) condition, which relates a model's capacity for effective long context length modeling to the scaling of its latent state size for storing past information. Our results are validated through experiments on both transformers and state space models. This work establishes a theoretical foundation that guides the development of large language models toward longer context lengths.

Denoising Hamiltonian Network for Physical Reasoning
Congyue Deng, Brandon Y. Feng, Cecilia Garraffo, Alan Garbarz, Robin Walters, William T. Freeman, Leonidas Guibas, Kaiming He
[ arXiv:2503.07596 ]
Foundational AI

Abstract Machine learning frameworks for physical problems must capture and enforce physical constraints that preserve the structure of dynamical systems. Many existing approaches achieve this by integrating physical operators into neural networks. While these methods offer theoretical guarantees, they face two key limitations: (i) they primarily model local relations between adjacent time steps, overlooking longer-range or higher-level physical interactions, and (ii) they focus on forward simulation while neglecting broader physical reasoning tasks. We propose the Denoising Hamiltonian Network (DHN), a novel framework that generalizes Hamiltonian mechanics operators into more flexible neural operators. DHN captures non-local temporal relationships and mitigates numerical integration errors through a denoising mechanism. DHN also supports multi-system modeling with a global conditioning mechanism. We demonstrate its effectiveness and flexibility across three diverse physical reasoning tasks with distinct inputs and outputs.

The Type I Superluminous Supernova Catalogue II: Spectroscopic Evolution in the Photospheric Phase, Velocity Measurements, and Constraints on Diversity
Aysha Aamer, Matt Nicholl, Sebastian Gomez, Edo Berger, Peter Blanchard, Joseph P. Anderson, Charlotte Angus, Amar Aryan, Chris Ashall, Ting-Wan Chen, Georgios Dimitriadis, Lluis Galbany, Anamaria Gkini, Mariusz Gromadzki, Claudia P. Gutierrez, Daichi Hiramatsu, Griffin Hosseinzadeh, Cosimo Inserra, Amit Kumar, Hanindyo Kuncarayakti, Giorgos Leloudas, Paolo Mazzali, Kyle Medler, Tomás E. Müller-Bravo, Mauricio Ramirez, Aiswarya Sankar.K, Steve Schulze, Avinash Singh, Jesper Sollerman, Shubham Srivastav, Jacco H. Terwel, David R. Young
[ arXiv:2503.21874 ]
Astrophysics

Abstract Hydrogen-poor superluminous supernovae (SLSNe) are among the most energetic explosions in the universe, reaching luminosities up to 100 times greater than those of normal supernovae. Detailed spectral analysis hold the potential to reveal their progenitors and underlying energy sources. This paper presents the largest compilation of SLSN photospheric spectra to date, encompassing data from ePESSTO+, the FLEET search and all published spectra up to December 2022. The dataset includes a total of 974 spectra of 234 SLSNe. By constructing average phase binned spectra, we find SLSNe initially exhibit high temperatures (10000 to 11000 K), with blue continua and weak lines. A rapid transformation follows, as temperatures drop to 5000 to 6000 K by 40 days post peak, leading to stronger P-Cygni features. These averages also suggest a fraction of SLSNe may contain some He at explosion. Variance within the dataset is slightly reduced when defining the phase of spectra relative to explosion, rather than peak, and normalising to the population's median e-folding time. Principal Component Analysis (PCA) supports this, requiring fewer components to explain the same level of variation when binning data by scaled days from explosion, suggesting a more homogeneous grouping. Using PCA and K-Means clustering, we identify outlying objects with unusual spectroscopic evolution and evidence for energy input from interaction, but find not support for groupings of two or more statistically significant subpopulations. We find Fe II {lambda}5169 lines velocities closely track the radius implied from blackbody fits, indicating formation near the photosphere. We also confirm a correlation between velocity and velocity gradient, which can be explained if all SLSNe are in homologous expansion but with different scale velocities. This behaviour aligns with expectations for an internal powering mechanism.

Decoding the Galactic Twirl: The Downfall of Milky Way-mass Galaxies Rotation Curves in the FIRE Simulations
Xiaowei Ou, Lina Necib, Andrew Wetzel, Anna Frebel, Jeremy Bailin, Micah Oeur
[ arXiv:2503.05877 ]
Astrophysics

Abstract Recent measurements of the Milky Way rotation curve found a sharp decline at around 15-20 kpc from the center of the Galaxy, suggesting that the Galactic dark matter halo is much less massive than predicted by other dynamical tracers. To address this tension, we study the validity of the assumptions made in calculating the Milky Way's rotation curve. To do so, we apply Jeans' equation, the current standard approach of measuring rotation curves, to three cosmological zoom-in simulations of Milky Way-like galaxies from the FIRE-2 Latte suite. Using synthetic Gaia surveys, we replicate the sample selection process and calculation employed in measuring the Milky Way rotation curve. We examine four failure modes of this calculation and find that the measured curves deviate from the true curve by 5-20% rather than below 5%, as estimated by previous works. Interestingly, there is a large galaxy-to-galaxy variance, and different systematics dominate different galaxies. We rederive the Milky Way's dark matter density profile with the rotation curve while incorporating systematics from the simulations. The posterior distribution of the density profiles is consistent with a fiducial NFW profile when assuming a gNFW profile for dark matter. We find that the virial mass, 7.32+1.98−1.53×1011 M⊙, consistent with other probes of the Milky Way's mass. However, we recommend that the field moves away from relying solely on the rotation curve when studying the dark matter profile, and adopts methods that incorporate additional probes and/or do not heavily depend on assumptions described in this study.

Trial by FIRE: Probing the dark matter density profile of dwarf galaxies with GraphNPE
Tri Nguyen, Justin Read, Lina Necib, Siddharth Mishra-Sharma, Claude-André Faucher-Giguère, Andrew Wetzel
[ arXiv:2503.03812 ]
Astrophysics

Abstract The Dark Matter (DM) distribution in dwarf galaxies provides crucial insights into both structure formation and the particle nature of DM. GraphNPE (Graph Neural Posterior Estimator), first introduced in Nguyen et al. (2023), is a novel simulation-based inference framework that combines graph neural networks and normalizing flows to infer the DM density profile from line-of-sight stellar velocities. Here, we apply GraphNPE to satellite dwarf galaxies in the FIRE-2 Latte simulation suite of Milky Way-mass halos, testing it against both Cold and Self-Interacting DM scenarios. Our method demonstrates superior precision compared to conventional Jeans-based approaches, recovering DM density profiles to within the 95% confidence level even in systems with as few as 30 tracers. Moreover, we present the first evaluation of mass modeling methods in constraining two key parameters from realistic simulations: the peak circular velocity, Vmax, and the peak virial mass, Mpeak200m. Using only line-of-sight velocities, GraphNPE can reliably recover both Vmax and Mpeak200m within our quoted uncertainties, including those experiencing tidal effects (≳ 63% of systems are recovered with our 68% confidence intervals and ≳ 92% within our 95% confidence intervals). The method achieves 10-20% accuracy in Vmax recovery, while Mpeak200m is recovered to 0.1-0.4 dex accuracy. This work establishes GraphNPE as a robust tool for inferring DM density profiles in dwarf galaxies, offering promising avenues for constraining DM models. The framework's potential extends beyond this study, as it can be adapted to non-spherical and disequilibrium models, showcasing the broader utility of simulation-based inference and graph-based learning in astrophysics.

Deriving Stellar Properties, Distances, and Reddenings using Photometry and Astrometry with BRUTUS
Joshua S. Speagle, Catherine Zucker, Angus Beane, Phillip A. Cargile, Aaron Dotter, Douglas P. Finkbeiner, Gregory M. Green, Benjamin D. Johnson, Edward F. Schlafly, Ana Bonaca, Charlie Conroy, Gwendolyn Eadie, Daniel J. Eisenstein, Alyssa A. Goodman, Jiwon Jesse Han, Harshil M. Kamdar, Rohan Naidu, Hans-Walter Rix, Andrew K. Saydjari, Yuan-Sen Ting, Ioana A. Zelko
[ arXiv:2503.02227 | code ]
Astrophysics

Abstract We present brutus, an open source Python package for quickly deriving stellar properties, distances, and reddenings to stars based on grids of stellar models constrained by photometric and astrometric data. We outline the statistical framework for deriving these quantities, its implementation, and various Galactic priors over the 3-D distribution of stars, stellar properties, and dust extinction (including RV variation). We establish a procedure to empirically calibrate MIST v1.2 isochrones by using open clusters to derive corrections to the effective temperatures and radii of the isochrones, which reduces systematic errors on the lower main sequence. We also describe and apply a method to estimate photometric offsets between stellar models and observed data using nearby, low-reddening field stars. We perform a series of tests on mock and real data to examine parameter recovery with MIST under different modeling assumptions, illustrating that brutus is able to recover distances and other stellar properties using optical to near-infrared photometry and astrometry.

A Deep, High-Angular Resolution 3D Dust Map of the Southern Galactic Plane
Catherine Zucker, Andrew K. Saydjari, Joshua S. Speagle, Edward F. Schlafly, Gregory M. Green, Robert Benjamin, Joshua Peek, Gordian Edenhofer, Alyssa Goodman, Michael A. Kuhn, Douglas P. Finkbeiner
[ arXiv:2503.02657 ]
Astrophysics

Abstract We present a deep, high-angular resolution 3D dust map of the southern Galactic plane over 239∘<ℓ<6∘ and |b|<10∘ built on photometry from the DECaPS2 survey, in combination with photometry from VVV, 2MASS, and unWISE and parallaxes from Gaia DR3 where available. To construct the map, we first infer the distance, extinction, and stellar types of over 700 million stars using the brutus stellar inference framework with a set of theoretical MIST stellar models. Our resultant 3D dust map has an angular resolution of 1′, roughly an order of magnitude finer than existing 3D dust maps and comparable to the angular resolution of the Herschel 2D dust emission maps. We detect complexes at the range of distances associated with the Sagittarius-Carina and Scutum-Centaurus arms in the fourth quadrant, as well as more distant structures out to a maximum reliable distance of d≈ 10 kpc from the Sun. The map is sensitive up to a maximum extinction of roughly AV≈12 mag. We publicly release both the stellar catalog and the 3D dust map, the latter of which can easily be queried via the Python package dustmaps. When combined with the existing Bayestar19 3D dust map of the northern sky, the DECaPS 3D dust map fills in the missing piece of the Galactic plane, enabling extinction corrections over the entire disk |b|<10∘. Our map serves as a pathfinder for the future of 3D dust mapping in the era of LSST and Roman, targeting regimes accessible with deep optical and near-infrared photometry but often inaccessible with Gaia.

Evidence for an Instability-Induced Binary Merger in the Double-Peaked, Helium-Rich Type IIn Supernova 2023zkd
A. Gagliano, V. A. Villar, T. Matsumoto, D. O. Jones, C. L. Ransome, A. E. Nugent, D. Hiramatsu, K. Auchettl, D. Tsuna, Y. Dong, S. Gomez, P. D. Aleo, C. Angus, T. de Boer, K. A. Bostroem, K. C. Chambers, D. A. Coulter, K. W. Davis, J. R. Fairlamb, J. Farah, D. Farias, R. J. Foley, C. Gall, H. Gao, E. P. Gonzalez, D. A. Howell, M. E. Huber, C. D. Kilpatrick, C.-C. Lin, M. E. MacLeod, E. A. Magnier, C. McCully, P. Minguez, G. Narayan, M. Newsome, K. C. Patra, A. Rest, S. Rest, S. Smartt, K. W. Smith, G. Terreran, R. J. Wainscoat, Q. Wang, S. K. Yadavalli, Y. Zenati (The Young Supernova Experiment)
[ arXiv:2502.19469 ]
Astrophysics

Abstract We present ultraviolet to infrared observations of the extraordinary Type IIn supernova 2023zkd (SN 2023zkd). Photometrically, it exhibits persistent and luminous precursor emission spanning ∼4 years preceding discovery (Mr≈−15 mag, 1,500~days in the observer frame), followed by a secondary stage of gradual brightening in its final year. Post-discovery, it exhibits two photometric peaks of comparable brightness (Mr≲−18.7 mag and Mr≈−18.4 mag, respectively) separated by 240 days. Spectroscopically, SN 2023zkd exhibits highly asymmetric and multi-component Balmer and He I profiles that we attribute to ejecta interaction with fast-moving (1,000−2,000kms−1) He-rich polar material and slow-moving (∼400kms−1) equatorially-distributed H-rich material. He II features also appear during the second light curve peak and evolve rapidly. Shock-driven models fit to the multi-band photometry suggest that the event is powered by interaction with ∼5−6M⊙ of CSM, with 2−3M⊙ associated with each light curve peak, expelled during mass-loss episodes ∼3−4 and ∼1−2 years prior to explosion. The observed precursor emission, combined with the extreme mass-loss rates required to power each light curve peak, favors either super-Eddington accretion onto a black hole or multiple long-lived eruptions from a massive star to luminosities that have not been previously observed. We consider multiple progenitor scenarios for SN 2023zkd, and find that the brightening optical precursor and inferred explosion properties are most consistent with a massive (MZAMS≥30M⊙) and partially-stripped He star undergoing an instability-induced merger with a black hole companion.

Seeing the Outer Edge of the Infant Type Ia Supernova 2024epr in the Optical and Near Infrared
W. B. Hoogendam, D. O. Jones, C. Ashall, B. J. Shappee, R. J. Foley, M. A. Tucker, M. E. Huber, K. Auchettl, D. D. Desai, A. Do, J. T. Hinkle, S. Romagnoli, J. Shi, A. Syncatto, C. R. Angus, K. C. Chambers, D. A. Coulter, K. W. Davis, T. de Boer, A. Gagliano, M. Kong, C.-C. Lin, T. B. Lowe, E. A. Magnier, P. Minguez, Y.-C. Pan, K.C. Patra, S. A. Severson, K. Taggart, A. R. Wasserman, S. K. Yadavalli
[ arXiv:2502.17556 ]
Astrophysics

Abstract We present optical-to-near infrared (NIR) photometry and spectroscopy of the Type Ia supernova (SN Ia) 2024epr, including NIR spectra observed within two days of first light. The early-time optical spectra show strong, high-velocity Ca and Si features near rarely-observed velocities at ∼0.1c, and the NIR spectra show a CI 'knee.' Despite these high-velocity features at early times, SN~2024epr evolves into a normal SN Ia, albeit with stronger peak-light Ca absorption than other SNe Ia with the same light curve shape. Although we infer a normal decline rate, Δm15(B)=1.09±0.12 mag, from the light-curve rise, SN 2024epr is a Branch 'cool' object and has red early-time colors (g−r≈0.15 mag at −10 days). The high velocities point to a density enhancement in the outer layers of the explosion, but thick-shell He-detonation models do not match the smoothly rising light curve or lack of He in our early-time NIR spectra. No current models (e.g., delayed detonation or thin He shell double detonation) appear to reproduce all of the observed properties. Such constraints are only possible for SN 2024epr from the earliest optical and NIR observations, highlighting their importance for constraining SN Ia models. Finally, we find several other SNe Ia with intermediate mass elements at ∼30,000 km s−1 within days after the explosion that evolve into otherwise normal SNe Ia at peak light, suggesting the early-time spectra of SNe Ia may hide a broad diversity of observational characteristics.

Anomaly preserving contrastive neural embeddings for end-to-end model-independent searches at the LHC
Kyle Metzger, Lana Xu, Mia Sodini, Thea K. Arrestad, Katya Govorkova, Gaia Grosso, Philip Harris
[ arXiv:2502.15926 ]
Experimental Physics

Abstract Anomaly detection -- identifying deviations from Standard Model predictions -- is a key challenge at the Large Hadron Collider due to the size and complexity of its datasets. This is typically addressed by transforming high-dimensional detector data into lower-dimensional, physically meaningful features. We tackle feature extraction for anomaly detection by learning powerful low-dimensional representations via contrastive neural embeddings. This approach preserves potential anomalies indicative of new physics and enables rare signal extraction using novel machine learning-based statistical methods for signal-independent hypothesis testing. We compare supervised and self-supervised contrastive learning methods, for both MLP- and Transformer-based neural embeddings, trained on the kinematic observables of physics objects in LHC collision events. The learned embeddings serve as input representations for signal-agnostic statistical detection methods in inclusive final states, achieving over ten fold improved detection performance over the original feature representation and up to four fold improvement over using a physics-informed selections of the same dimensionality. We achieve significant improvement in discovery power for both rare new physics signals and rare Standard Model processes across diverse final states, demonstrating its applicability for efficiently searching for diverse signals simultaneously. We show that the optimal representation for background classification does not always maximize sensitivity to new physics signals, revealing an inherent trade-off between background structure preservation and anomaly enhancement. Our findings demonstrate that foundation models for particle physics data hold significant potential for improving neural feature extraction, enabling scientific discovery in inclusive final states at collider experiments.

Isolating Unisolated Upsilons with Anomaly Detection in CMS Open Data
Rikab Gambhir, Radha Mastandrea, Benjamin Nachman, Jesse Thaler
[ arXiv:2502.14036 | code ]
Experimental Physics

Abstract We present the first study of anti-isolated Upsilon decays to two muons (Υ→μ+μ−) in proton-proton collisions at the Large Hadron Collider. Using a machine learning (ML)-based anomaly detection strategy, we 'rediscover' the Υ in 13 TeV CMS Open Data from 2016, despite overwhelming anti-isolated backgrounds. We elevate the signal significance to 6.4σ using these methods, starting from 1.6σ using the dimuon mass spectrum alone. Moreover, we demonstrate improved sensitivity from using an ML-based estimate of the multi-feature likelihood compared to traditional 'cut-and-count' methods. Our work demonstrates that it is possible and practical to find real signals in experimental collider data using ML-based anomaly detection, and we distill a readily-accessible benchmark dataset from the CMS Open Data to facilitate future anomaly detection developments.

On the Learnability of Knot Invariants: Representation, Predictability, and Neural Similarity
Audrey Lindsay, Fabian Ruehle
[ arXiv:2502.12243 ]
Foundational AI

Abstract We analyze different aspects of neural network predictions of knot invariants. First, we investigate the impact of different knot representations on the prediction of invariants and find that braid representations work in general the best. Second, we study which knot invariants are easy to learn, with invariants derived from hyperbolic geometry and knot diagrams being very easy to learn, while invariants derived from topological or homological data are harder. Predicting the Arf invariant could not be learned for any representation. Third, we propose a cosine similarity score based on gradient saliency vectors, and a joint misclassification score to uncover similarities in neural networks trained to predict related topological invariants.

Interpretable Machine Learning for Kronecker Coefficients
Giorgi Butbaia, Kyu-Hwan Lee, Fabian Ruehle
[ arXiv:2502.11774 ]
Foundational AI

Abstract We analyze the saliency of neural networks and employ interpretable machine learning models to predict whether the Kronecker coefficients of the symmetric group are zero or not. Our models use triples of partitions as input features, as well as b-loadings derived from the principal component of an embedding that captures the differences between partitions. Across all approaches, we achieve an accuracy of approximately 83% and derive explicit formulas for a decision function in terms of b-loadings. Additionally, we develop transformer-based models for prediction, achieving the highest reported accuracy of over 99%.

Bridging the Sim-to-Real Gap for Athletic Loco-Manipulation
Nolan Fey, Gabriel B. Margolis, Martin Peticco, Pulkit Agrawal
[ arXiv:2502.10894 ]
Foundational AI

Abstract Achieving athletic loco-manipulation on robots requires moving beyond traditional tracking rewards - which simply guide the robot along a reference trajectory - to task rewards that drive truly dynamic, goal-oriented behaviors. Commands such as 'throw the ball as far as you can' or 'lift the weight as quickly as possible' compel the robot to exhibit the agility and power inherent in athletic performance. However, training solely with task rewards introduces two major challenges: these rewards are prone to exploitation (reward hacking), and the exploration process can lack sufficient direction. To address these issues, we propose a two-stage training pipeline. First, we introduce the Unsupervised Actuator Net (UAN), which leverages real-world data to bridge the sim-to-real gap for complex actuation mechanisms without requiring access to torque sensing. UAN mitigates reward hacking by ensuring that the learned behaviors remain robust and transferable. Second, we use a pre-training and fine-tuning strategy that leverages reference trajectories as initial hints to guide exploration. With these innovations, our robot athlete learns to lift, throw, and drag with remarkable fidelity from simulation to reality.

LEAPS: A discrete neural sampler via locally equivariant networks
Peter Holderrieth, Michael S. Albergo, Tommi Jaakkola
[ arXiv:2502.10843 ]
Foundational AI

Abstract We propose LEAPS, an algorithm to sample from discrete distributions known up to normalization by learning a rate matrix of a continuous-time Markov chain (CTMC). LEAPS can be seen as a continuous-time formulation of annealed importance sampling and sequential Monte Carlo methods, extended so that the variance of the importance weights is offset by the inclusion of the CTMC. To derive these importance weights, we introduce a set of Radon-Nikodym derivatives of CTMCs over their path measures. Because the computation of these weights is intractable with standard neural network parameterizations of rate matrices, we devise a new compact representation for rate matrices via what we call locally equivariant functions. To parameterize them, we introduce a family of locally equivariant multilayer perceptrons, attention layers, and convolutional networks, and provide an approach to make deep networks that preserve the local equivariance. This property allows us to propose a scalable training algorithm for the rate matrix such that the variance of the importance weights associated to the CTMC are minimal. We demonstrate the efficacy of LEAPS on problems in statistical physics.

Do We Need to Verify Step by Step? Rethinking Process Supervision from a Theoretical Perspective
Zeyu Jia, Alexander Rakhlin, Tengyang Xie
[ arXiv:2502.10581 ]
Foundational AI

Abstract As large language models have evolved, it has become crucial to distinguish between process supervision and outcome supervision -- two key reinforcement learning approaches to complex reasoning tasks. While process supervision offers intuitive advantages for long-term credit assignment, the precise relationship between these paradigms has remained an open question. Conventional wisdom suggests that outcome supervision is fundamentally more challenging due to the trajectory-level coverage problem, leading to significant investment in collecting fine-grained process supervision data. In this paper, we take steps towards resolving this debate. Our main theorem shows that, under standard data coverage assumptions, reinforcement learning through outcome supervision is no more statistically difficult than through process supervision, up to polynomial factors in horizon. At the core of this result lies the novel Change of Trajectory Measure Lemma -- a technical tool that bridges return-based trajectory measure and step-level distribution shift. Furthermore, for settings with access to a verifier or a rollout capability, we prove that any policy's advantage function can serve as an optimal process reward model, providing a direct connection between outcome and process supervision. These findings suggest that the empirically observed performance gap -- if any -- between outcome and process supervision likely stems from algorithmic limitations rather than inherent statistical difficulties, potentially transforming how we approach data collection and algorithm design for reinforcement learning.

Machine learning the vanishing order of rational L-functions
Joanna Bieri, Giorgi Butbaia, Edgar Costa, Alyson Deines, Kyu-Hwan Lee, David Lowry-Duda, Thomas Oliver, Yidi Qi, Tamara Veenstra
[ arXiv:2502.10360 ]
Theoretical Physics

Abstract In this paper, we study the vanishing order of rational L-functions from a data scientific perspective. Each L-function is represented in our data by finitely many Dirichlet coefficients, the normalisation of which depends on the context. We observe murmuration-like patterns in averages across our dataset, find that PCA clusters rational L-functions by their vanishing order, and record that LDA and neural networks may accurately predict this quantity.

Parameter Symmetry Breaking and Restoration Determines the Hierarchical Learning in AI Systems
Liu Ziyin, Yizhou Xu, Tomaso Poggio, Isaac Chuang
[ arXiv:2502.05300 ]
Foundational AI

Abstract The dynamics of learning in modern large AI systems is hierarchical, often characterized by abrupt, qualitative shifts akin to phase transitions observed in physical systems. While these phenomena hold promise for uncovering the mechanisms behind neural networks and language models, existing theories remain fragmented, addressing specific cases. In this paper, we posit that parameter symmetry breaking and restoration serve as a unifying mechanism underlying these behaviors. We synthesize prior observations and show how this mechanism explains three distinct hierarchies in neural networks: learning dynamics, model complexity, and representation formation. By connecting these hierarchies, we highlight symmetry -- a cornerstone of theoretical physics -- as a potential fundamental principle in modern AI.

Harmonic Loss Trains Interpretable AI Models
David D. Baek, Ziming Liu, Riya Tyagi, Max Tegmark
[ arXiv:2502.01628 ]
Foundational AI

Abstract In this paper, we introduce **harmonic loss** as an alternative to the standard cross-entropy loss for training neural networks and large language models (LLMs). Harmonic loss enables improved interpretability and faster convergence, owing to its scale invariance and finite convergence point by design, which can be interpreted as a class center. We first validate the performance of harmonic models across algorithmic, vision, and language datasets. Through extensive experiments, we demonstrate that models trained with harmonic loss outperform standard models by: (a) enhancing interpretability, (b) requiring less data for generalization, and (c) reducing grokking. Moreover, we compare a GPT-2 model trained with harmonic loss to the standard GPT-2, illustrating that the harmonic model develops more interpretable representations. Looking forward, we believe harmonic loss has the potential to become a valuable tool in domains with limited data availability or in high-stakes applications where interpretability and reliability are paramount, paving the way for more robust and efficient neural network models.

A Poisson Process AutoDecoder for X-ray Sources
Yanke Song, Victoria Ashley Villar, Juan Rafael Martinez-Galarza, Steven Dillmann
[ arXiv:2502.01627 ]
Astrophysics

Abstract X-ray observing facilities, such as the Chandra X-ray Observatory and the eROSITA, have detected millions of astronomical sources associated with high-energy phenomena. The arrival of photons as a function of time follows a Poisson process and can vary by orders-of-magnitude, presenting obstacles for common tasks such as source classification, physical property derivation, and anomaly detection. Previous work has either failed to directly capture the Poisson nature of the data or only focuses on Poisson rate function reconstruction. In this work, we present Poisson Process AutoDecoder (PPAD). PPAD is a neural field decoder that maps fixed-length latent features to continuous Poisson rate functions across energy band and time via unsupervised learning. PPAD reconstructs the rate function and yields a representation at the same time. We demonstrate the efficacy of PPAD via reconstruction, regression, classification and anomaly detection experiments using the Chandra Source Catalog.

Language Models Use Trigonometry to Do Addition
Subhash Kantamneni, Max Tegmark
[ arXiv:2502.00873 ]
Foundational AI

Abstract Mathematical reasoning is an increasingly important indicator of large language model (LLM) capabilities, yet we lack understanding of how LLMs process even simple mathematical tasks. To address this, we reverse engineer how three mid-sized LLMs compute addition. We first discover that numbers are represented in these LLMs as a generalized helix, which is strongly causally implicated for the tasks of addition and subtraction, and is also causally relevant for integer division, multiplication, and modular arithmetic. We then propose that LLMs compute addition by manipulating this generalized helix using the 'Clock' algorithm: to solve a+b, the helices for a and b are manipulated to produce the a+b answer helix which is then read out to model logits. We model influential MLP outputs, attention head outputs, and even individual neuron preactivations with these helices and verify our understanding with causal interventions. By demonstrating that LLMs represent numbers on a helix and manipulate this helix to perform addition, we present the first representation-level explanation of an LLM's mathematical capability.

Progress in Normalizing Flows for 4d Gauge Theories
Ryan Abbott, Denis Boyda, Daniel C. Hackett, Gurtej Kanwar, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2502.00263 ]
Theoretical Physics

Abstract Normalizing flows have arisen as a tool to accelerate Monte Carlo sampling for lattice field theories. This work reviews recent progress in applying normalizing flows to 4-dimensional nonabelian gauge theories, focusing on two advancements: an architectural improvement referred to as learned active loops, and the application of correlated ensemble methods to QCD with Nf=2 dynamical fermions.

A Fast Periodicity Detection Algorithm Sensitive to Arbitrary Waveforms
Douglas P. Finkbeiner, Thomas A. Prince, Samuel E. Whitebook
[ arXiv:2502.00243 ]
Astrophysics

Abstract A reexamination of period finding algorithms is prompted by new large area astronomical sky surveys that can identify billions of individual sources having a thousand or more observations per source. This large increase in data necessitates fast and efficient period detection algorithms. In this paper, we provide an initial description of an algorithm that is being used for detection of periodic behavior in a sample of 1.5 billion objects using light curves generated from Zwicky Transient Facility (ZTF) data (Bellm et al. 2019; Masci et al. 2018). We call this algorithm 'Fast Periodicity Weighting' (FPW), derived using a Gaussian Process (GP) formalism. A major advantage of the FPW algorithm for ZTF analysis is that it is agnostic to the details of the phase-folded waveform. Periodic sources in ZTF show a wide variety of waveforms, some quite complex, including eclipsing objects, sinusoidally varying objects also exhibiting eclipses, objects with cyclotron emission at various phases, and accreting objects with complex waveforms. We describe the FPW algorithm and its application to ZTF, and provide efficient code for both CPU and GPU.

Low-Rank Adapting Models for Sparse Autoencoders
Matthew Chen, Joshua Engels, Max Tegmark
[ arXiv:2501.19406 | code ]
Foundational AI

Abstract Sparse autoencoders (SAEs) decompose language model representations into a sparse set of linear latent vectors. Recent works have improved SAEs using language model gradients, but these techniques require many expensive backward passes during training and still cause a significant increase in cross entropy loss when SAE reconstructions are inserted into the model. In this work, we improve on these limitations by taking a fundamentally different approach: we use low-rank adaptation (LoRA) to finetune the language model itself around a previously trained SAE. We analyze our method across SAE sparsity, SAE width, language model size, LoRA rank, and model layer on the Gemma Scope family of SAEs. In these settings, our method reduces the cross entropy loss gap by 30% to 55% when SAEs are inserted during the forward pass. We also find that compared to end-to-end (e2e) SAEs, our approach achieves the same downstream cross entropy loss 3× to 20× faster on Gemma-2-2B and 2× to 10× faster on Llama-3.2-1B. We further show that our technique improves downstream metrics and can adapt multiple SAEs at once. Our results demonstrate that improving model interpretability is not limited to post-hoc SAE training; Pareto improvements can also be achieved by directly optimizing the model itself.

Central densities of dark matter halos in FIRE-2 simulations of low-mass galaxies with cold dark matter and self-interacting dark matter
Maria C. Straight, Michael Boylan-Kolchin, James S. Bullock, Philip F. Hopkins, Xuejian Shen, Lina Necib, Alexandres Lazar, Andrew S. Graus, Jenna Samuel
[ arXiv:2501.16602 ]
Astrophysics

Abstract We investigate the central density structure of dark matter halos in cold dark matter (CDM) and self-interacting dark matter (SIDM) models using simulations that are part of the Feedback In Realistic Environments (FIRE) project. For simulated halos of dwarf galaxy scale (Mhalo(z=0)≈1010M⊙), we study the central structure in both dissipationless simulations and simulations with full FIRE-2 galaxy formation physics. As has been demonstrated extensively in recent years, both baryonic feedback and self-interactions can convert central cusps into cores, with the former process doing so in a manner that depends sensitively on stellar mass at fixed Mhalo. Whether the two processes (baryonic feedback and self-interactions) are distinguishable, however, remains an open question. Here we demonstrate that, compared to feedback-induced cores, SIDM-induced cores transition more quickly from the central region of constant density to the falling density at larger radial scales. This result holds true even when including identical galaxy formation modeling in SIDM simulations as is used in CDM simulations, since self-interactions dominate over galaxy formation physics in establishing the central structure of SIDM halos in this mass regime. The change in density profile slope as a function of radius therefore holds the potential to discriminate between self-interactions and galaxy formation physics as the driver of core formation in dwarf galaxies.

QCD Theory meets Information Theory
Benoît Assi, Stefan Höche, Kyle Lee, Jesse Thaler
[ arXiv:2501.17219 ]
Experimental Physics

Abstract We present a novel technique to incorporate precision calculations from quantum chromodynamics into fully differential particle-level Monte-Carlo simulations. By minimizing an information-theoretic quantity subject to constraints, we achieve consistency with the theory input and its estimated systematic uncertainties. Our method can be applied to arbitrary observables known from precision calculations, including multiple observables simultaneously. It generates strictly positive weights, thus offering a clear path to statistically powerful and theoretically precise computations for current and future collider experiments. As a proof of concept, we apply our technique to event-shape observables at electron-positron colliders, leveraging existing precision calculations of thrust. Our analysis highlights the importance of logarithmic moments of event shapes, which have not been previously studied in the collider physics literature.

A theoretical approach to density-split clustering
Mathilde Pinon, Arnaud de Mattia, Étienne Burtin, Vanina Ruhlmann-Kleider, Sandrine Codis, Enrique Paillas, Carolina Cuesta-Lazaro
[ arXiv:2501.14638 | code ]
Astrophysics

Abstract We present an analytical model for density-split correlation functions, that probe galaxy clustering in different density environments. Specifically, we focus on the cross-correlation between density-split regions and the tracer density field. We show that these correlation functions can be expressed in terms of the two-point probability density function (PDF) of the density field, or equivalently, its bias function. We derive analytical predictions using three levels of approximation for the two-point PDF: a bivariate Gaussian distribution, a bivariate shifted log-normal distribution, and a prediction based on the Large Deviation Theory framework. Under spherical symmetry – where spherical collapse dynamics apply, such as for count-in-cell densities with spherical top-hat smoothing – LDT predicts the density two-point PDF in the large-separation regime relative to the smoothing radius. We validate our model against dark matter N-body simulations in real space, incorporating Poisson shot noise. Our results show that the LDT predictions outperform the log-normal approximation, and agrees with simulations on large scales within the cosmic variance of a typical DESI DR1 sample, despite relying on only one degree of freedom.

Theoretical Predictions for the Inner Dark Matter Distribution in the Milky Way Informed by Simulations
Abdelaziz Hussein, Lina Necib, Manoj Kaplinghat, Stacy Y. Kim, Andrew Wetzel, Justin I. Read, Martin P. Rey, Oscar Agertz
[ arXiv:2501.14868 | code ]
Astrophysics

Abstract We build a theoretical range for the Milky Way's (MW) inner dark matter (DM) distribution informed by the FIRE-2, Auriga, VINTERGATAN-GM, and TNG50 simulation suites assuming the canonical cold dark matter (CDM) model. The DM density profiles in Auriga, VINTERGATAN-GM, and TNG50 can be approximately modeled using the adiabatic contraction prescription of Gnedin et al. 2004, while FIRE-2 has stronger baryonic feedback, leading to a departure from the adiabatic contraction model. The simulated halos that are adiabatically contracted are close to spherical (axis ratio q∈[0.75−0.9] at 5∘), whereas halos that experience strong baryonic feedback are oblate (q∈[0.5−0.7]). Using the adiabatic contraction and strong baryonic feedback models, along with the observed stellar distribution of the MW, the inner logarithmic density slope for CDM in the MW is predicted to range from −0.5 to −1.3. The J-factor, which determines the DM-annihilation flux, averaged over a solid angle of 5∘ (10∘) is predicted to span the range 0.8-30 (0.6-10) ×1023GeV2/cm5. The D-factor, which determines the flux due to DM decay, is predicted to be in the range 0.6-2 (0.5−1) ×1023GeV/cm2.

SIDDA: SInkhorn Dynamic Domain Adaptation for Image Classification with Equivariant Neural Networks
Sneh Pandya, Purvik Patel, Brian D. Nord, Mike Walmsley, Aleksandra Ćiprijanović
[ arXiv:2501.14048 | code ]
Foundational AI

Abstract Modern neural networks (NNs) often do not generalize well in the presence of a 'covariate shift'; that is, in situations where the training and test data distributions differ, but the conditional distribution of classification labels remains unchanged. In such cases, NN generalization can be reduced to a problem of learning more domain-invariant features. Domain adaptation (DA) methods include a range of techniques aimed at achieving this; however, these methods have struggled with the need for extensive hyperparameter tuning, which then incurs significant computational costs. In this work, we introduce SIDDA, an out-of-the-box DA training algorithm built upon the Sinkhorn divergence, that can achieve effective domain alignment with minimal hyperparameter tuning and computational overhead. We demonstrate the efficacy of our method on multiple simulated and real datasets of varying complexity, including simple shapes, handwritten digits, and real astronomical observations. SIDDA is compatible with a variety of NN architectures, and it works particularly well in improving classification accuracy and model calibration when paired with equivariant neural networks (ENNs). We find that SIDDA enhances the generalization capabilities of NNs, achieving up to a ≈40% improvement in classification accuracy on unlabeled target data. We also study the efficacy of DA on ENNs with respect to the varying group orders of the dihedral group DN, and find that the model performance improves as the degree of equivariance increases. Finally, we find that SIDDA enhances model calibration on both source and target data--achieving over an order of magnitude improvement in the ECE and Brier score. SIDDA's versatility, combined with its automated approach to domain alignment, has the potential to advance multi-dataset studies by enabling the development of highly generalizable models.

Physics of Skill Learning
Ziming Liu, Yizhou Liu, Eric J. Michaud, Jeff Gore, Max Tegmark
[ arXiv:2501.12391 | code ]
Foundational AI

Abstract We aim to understand physics of skill learning, i.e., how skills are learned in neural networks during training. We start by observing the Domino effect, i.e., skills are learned sequentially, and notably, some skills kick off learning right after others complete learning, similar to the sequential fall of domino cards. To understand the Domino effect and relevant behaviors of skill learning, we take physicists' approach of abstraction and simplification. We propose three models with varying complexities -- the Geometry model, the Resource model, and the Domino model, trading between reality and simplicity. The Domino effect can be reproduced in the Geometry model, whose resource interpretation inspires the Resource model, which can be further simplified to the Domino model. These models present different levels of abstraction and simplification; each is useful to study some aspects of skill learning. The Geometry model provides interesting insights into neural scaling laws and optimizers; the Resource model sheds light on the learning dynamics of compositional tasks; the Domino model reveals the benefits of modularity. These models are not only conceptually interesting -- e.g., we show how Chinchilla scaling laws can emerge from the Geometry model, but also are useful in practice by inspiring algorithmic development -- e.g., we show how simple algorithmic changes, motivated by these toy models, can speed up the training of deep learning models.

Average-Reward Reinforcement Learning with Entropy Regularization
Jacob Adamczyk, Volodymyr Makarenko, Stas Tiomkin, Rahul V. Kulkarni
[ arXiv:2501.09080 | code ]
Foundational AI

Abstract The average-reward formulation of reinforcement learning (RL) has drawn increased interest in recent years due to its ability to solve temporally-extended problems without discounting. Independently, RL algorithms have benefited from entropy-regularization: an approach used to make the optimal policy stochastic, thereby more robust to noise. Despite the distinct benefits of the two approaches, the combination of entropy regularization with an average-reward objective is not well-studied in the literature and there has been limited development of algorithms for this setting. To address this gap in the field, we develop algorithms for solving entropy-regularized average-reward RL problems with function approximation. We experimentally validate our method, comparing it with existing algorithms on standard benchmarks for RL.

Inferring Transition Dynamics from Value Functions
Jacob Adamczyk
[ arXiv:2501.09081 ]
Foundational AI

Abstract In reinforcement learning, the value function is typically trained to solve the Bellman equation, which connects the current value to future values. This temporal dependency hints that the value function may contain implicit information about the environment's transition dynamics. By rearranging the Bellman equation, we show that a converged value function encodes a model of the underlying dynamics of the environment. We build on this insight to propose a simple method for inferring dynamics models directly from the value function, potentially mitigating the need for explicit model learning. Furthermore, we explore the challenges of next-state identifiability, discussing conditions under which the inferred dynamics model is well-defined. Our work provides a theoretical foundation for leveraging value functions in dynamics modeling and opens a new avenue for bridging model-free and model-based reinforcement learning.

An Updated Detection Pipeline for Precursor Emission in Type II Supernova 2020tlf
Wynn Jacobson-Galán, Sebastian Gonzalez, Shreyas Patel, Luc Dessart, David Jones, Deanne Coppejans, Georgios Dimitriadis, Ryan J. Foley, Charles D. Kilpatrick, David Matthews, Sofia Rest, Giacomo Terreran, Patrick D. Aleo, Katie Auchettl, Peter K. Blanchard, David A. Coulter, Kyle W. Davis, Thomas de Boer, Lindsay DeMarchi, Maria R. Drout, Nicholas Earl, Alexander Gagliano, Christa Gall, Jens Hjorth, Mark E. Huber, Adaeze L. Ibik, Danny Milisavljevic, Yen-Chen Pan, Armin Rest, Ryan Ridden-Harper, Cesar Rojas-Bravo, Matthew R. Siebert, Ken W. Smith, Kirsty Taggart, Samaporn Tinyanont, Qinan Wang, Yossef Zenati
The American Astronomical Society, Volume 9, Number 1, 2025 [ arXiv:2501.08475 ] published
Astrophysics

Abstract We present a new photometric pipeline for the detection of pre-supernova (pre-SN) emission in the Young Supernova Experiment (YSE) sky survey. The method described is applied to SN 2020tlf, a type II SN (SN II) with precursor emission in the last ~100 days before first light. We re-analyze the YSE griz-band light curves of SN 2020tlf and provide revised pre-explosion photometry that includes a robust list of confident detection and limiting magnitudes. Compared to the results of Jacobson-Galan et al. 2022a, this new analysis yields fewer total r/i/z-band pre-SN detections at phases > -100 days. Furthermore, we discourage the use of the blackbody modeling of the pre-explosion spectral energy distribution, the pre-SN bolometric light curve and the blackbody model parameters presented in Jacobson-Galan et al. 2022a. Nevertheless, binned photometry of SN 2020tlf confirms a consistent progenitor luminosity of ~1040 erg s−1 before explosion.

Lake- and Surface-Based Detectors for Forward Neutrino Physics
Nicholas W. Kamp, Carlos A. Argüelles, Albrecht Karle, Jennifer Thomas, Tianlu Yuan
[ arXiv:2501.08278 ]
Experimental Physics

Abstract We propose two medium-baseline, kiloton-scale neutrino experiments to study neutrinos from LHC proton-proton collisions: SINE, a surface-based scintillator panel detector observing muon neutrinos from the CMS interaction point, and UNDINE, a water Cherenkov detector submerged in lake Geneva observing all-flavor neutrinos from LHCb. Using a Monte Carlo simulation, we estimate millions of neutrino interactions during the high-luminosity LHC era. We show that these datasets can constrain neutrino cross sections, charm production in pp collisions, and strangeness enhancement as a solution to the cosmic-ray muon puzzle. SINE and UNDINE thus offer a cost-effective medium-baseline complement to the proposed short-baseline forward physics facility.

Synthesis and Analysis of Data as Probability Measures with Entropy-Regularized Optimal Transport
Brendan Mallery, James M. Murphy, Shuchin Aeron
[ arXiv:2501.07446 | code ]
Foundational AI

Abstract We consider synthesis and analysis of probability measures using the entropy-regularized Wasserstein-2 cost and its unbiased version, the Sinkhorn divergence. The synthesis problem consists of computing the barycenter, with respect to these costs, of m reference measures given a set of coefficients belonging to the m-dimensional simplex. The analysis problem consists of finding the coefficients for the closest barycenter in the Wasserstein-2 distance to a given measure μ. Under the weakest assumptions on the measures thus far in the literature, we compute the derivative of the entropy-regularized Wasserstein-2 cost. We leverage this to establish a characterization of regularized barycenters as solutions to a fixed-point equation for the average of the entropic maps from the barycenter to the reference measures. This characterization yields a finite-dimensional, convex, quadratic program for solving the analysis problem when μ is a barycenter. It is shown that these coordinates, as well as the value of the barycenter functional, can be estimated from samples with dimension-independent rates of convergence, a hallmark of entropy-regularized optimal transport, and we verify these rates experimentally. We also establish that barycentric coordinates are stable with respect to perturbations in the Wasserstein-2 metric, suggesting a robustness of these coefficients to corruptions. We employ the barycentric coefficients as features for classification of corrupted point cloud data, and show that compared to neural network baselines, our approach is more efficient in small training data regimes.

Effects of galactic environment on size and dark matter content in low-mass galaxies
Francisco J. Mercado, Jorge Moreno, Robert Feldmann, Marckie Zeender, Jose A. Benavides, Joanna M. Piotrowska, Courtney Klein, Coral Wheeler, Lina Necib, James S. Bullock, Philip F. Hopkins
[ arXiv:2501.04084 ]
Astrophysics

Abstract We utilize the cosmological volume simulation, FIREbox, to investigate how a galaxy's environment influences its size and dark matter content. Our study focuses on approximately 1,200 galaxies (886 central and 332 satellite halos) in the low-mass regime, with stellar masses between 106 to 109 M⊙. We analyze the size-mass relation (r50−M⋆), inner dark matter mass-stellar mass (M50DM−M⋆) relation, and the halo mass-stellar mass (Mhalo−M⋆) relation. At fixed stellar mass, we find the galaxies experiencing stronger tidal influences, indicated by higher Perturbation Indices (PI > 1) are generally larger and have lower masses relative to their counterparts with lower Perturbation Indices (PI < 1). Applying a Random Forest regression model, we show that both the environment (PI) and halo mass (Mrmhalo) are significant predictors of a galaxy's relative size and dark matter content. Notably, because Mhalo is also strongly affected by the environment, our findings indicate that environmental conditions not only influence galactic sizes and relative inner dark matter content directly, but also indirectly through their impact on halo mass. Our results highlight a critical interplay between environmental factors and halo mass in shaping galaxy properties, affirming the environment as a fundamental driver in galaxy formation and evolution.

Predicting band gap from chemical composition: A simple learned model for a material property with atypical statistics
Andrew Ma, Owen Dugan, Marin Soljačić
[ arXiv:2501.02932 ]
Foundational AI

Abstract In solid-state materials science, substantial efforts have been devoted to the calculation and modeling of the electronic band gap. While a wide range of ab initio methods and machine learning algorithms have been created that can predict this quantity, the development of new computational approaches for studying the band gap remains an active area of research. Here we introduce a simple machine learning model for predicting the band gap using only the chemical composition of the crystalline material. To motivate the form of the model, we first analyze the empirical distribution of the band gap, which sheds new light on its atypical statistics. Specifically, our analysis enables us to frame band gap prediction as a task of modeling a mixed random variable, and we design our model accordingly. Our model formulation incorporates thematic ideas from chemical heuristic models for other material properties in a manner that is suited towards the band gap modeling task. The model has exactly one parameter corresponding to each element, which is fit using data. To predict the band gap for a given material, the model computes a weighted average of the parameters associated with its constituent elements and then takes the maximum of this quantity and zero. The model provides heuristic chemical interpretability by intuitively capturing the associations between the band gap and individual chemical elements.

Exoplanet Detection via Differentiable Rendering
Brandon Y. Feng, Rodrigo Ferrer-Chávez, Aviad Levis, Jason J. Wang, Katherine L. Bouman, William T. Freeman
IEEE Transactions on Computational Imaging, 2025, Volume 11 [ arXiv:2501.01912 ] published
Astrophysics

Abstract Direct imaging of exoplanets is crucial for advancing our understanding of planetary systems beyond our solar system, but it faces significant challenges due to the high contrast between host stars and their planets. Wavefront aberrations introduce speckles in the telescope science images, which are patterns of diffracted starlight that can mimic the appearance of planets, complicating the detection of faint exoplanet signals. Traditional post-processing methods, operating primarily in the image intensity domain, do not integrate wavefront sensing data. These data, measured mainly for adaptive optics corrections, have been overlooked as a potential resource for post-processing, partly due to the challenge of the evolving nature of wavefront aberrations. In this paper, we present a differentiable rendering approach that leverages these wavefront sensing data to improve exoplanet detection. Our differentiable renderer models wave-based light propagation through a coronagraphic telescope system, allowing gradient-based optimization to significantly improve starlight subtraction and increase sensitivity to faint exoplanets. Simulation experiments based on the James Webb Space Telescope configuration demonstrate the effectiveness of our approach, achieving substantial improvements in contrast and planet detection limits. Our results showcase how the computational advancements enabled by differentiable rendering can revitalize previously underexploited wavefront data, opening new avenues for enhancing exoplanet imaging and characterization.

Cosmological constraints from the Minkowski functionals of the BOSS CMASS galaxy sample
Wei Liu, Enrique Paillas, Carolina Cuesta-Lazaro, Georgios Valogiannis, Wenjuan Fang
[ arXiv:2501.01698 ]
Astrophysics

Abstract For the first time, we develop a simulation-based model for the Minkowski functionals (MFs) of large-scale structure, which allows us to extract the full information available from the MFs (including both the Gaussian and non-Gaussian part), and apply it to the BOSS DR12 CMASS galaxy sample. Our model is based on high-fidelity mock galaxy catalogs constructed from the extsc{Abacus} extsc{Summit} simulations using the halo occupation distribution (HOD) framework, which include the redshift-space distortions and Alcock-Paczynski distortions, incorporate survey realism, including survey geometry and veto masks, and account for angular plus radial selection effects. The cosmological and HOD parameter dependence of the MFs is captured with a neural network emulator trained from the galaxy mocks with various cosmological and HOD parameters. To benchmark the constraining power of the MFs, we also train an emulator for the galaxy 2-point correlation function (2PCF) using the same pipeline. Having validated our approach through successful parameter recovery tests on both internal and external mocks, including non-HOD forward models of the halo-galaxy connection, we apply our forward model to analyze the CMASS data in the redshift range 0.45<z<0.58. We find the MFs provide stronger constraints on the cosmological parameters than the 2PCF. The combination of the two gives ωcdm=0.1172+0.0020−0.0023, σ8=0.783±0.026, and ns=0.966+0.019−0.015, which are tighter by a factor of 2.0, 1.9, and 1.6 than the 2PCF alone. The derived constraint fσ8=0.453±0.016 is also improved by a factor of 1.9, compared to the 2PCF, and agrees well with Planck 2018 predictions and other results from a series of studies in the literature.

Robust resonant anomaly detection with NPLM
Gaia Grosso, Debajyoti Sengupta, Tobias Golling, Philip Harris
[ arXiv:2501.01778 ]
Experimental Physics

Abstract In this study, we investigate the application of the New Physics Learning Machine (NPLM) algorithm as an alternative to the standard CWoLa method with Boosted Decision Trees (BDTs), particularly for scenarios with rare signal events. NPLM offers an end-to-end approach to anomaly detection and hypothesis testing by utilizing an in-sample evaluation of a binary classifier to estimate a log-density ratio, which can improve detection performance without prior assumptions on the signal model. We examine two approaches: (1) a end-to-end NPLM application in cases with reliable background modelling and (2) an NPLM-based classifier used for signal selection when accurate background modelling is unavailable, with subsequent performance enhancement through a hyper-test on multiple values of the selection threshold. Our findings show that NPLM-based methods outperform BDT-based approaches in detection performance, particularly in low signal injection scenarios, while significantly reducing epistemic variance due to hyperparameter choices. This work highlights the potential of NPLM for robust resonant anomaly detection in particle physics, setting a foundation for future methods that enhance sensitivity and consistency under signal variability.

Learning Fricke signs from Maass form Coefficients
Joanna Bieri, Giorgi Butbaia, Edgar Costa, Alyson Deines, Kyu-Hwan Lee, David Lowry-Duda, Thomas Oliver, Yidi Qi, Tamara Veenstra
[ arXiv:2501.02105 ]
Theoretical Physics

Abstract In this paper, we conduct a data-scientific investigation of Maass forms. We find that averaging the Fourier coefficients of Maass forms with the same Fricke sign reveals patterns analogous to the recently discovered 'murmuration' phenomenon, and that these patterns become more pronounced when parity is incorporated as an additional feature. Approximately 43% of the forms in our dataset have an unknown Fricke sign. For the remaining forms, we employ Linear Discriminant Analysis (LDA) to machine learn their Fricke sign, achieving 96% (resp. 94%) accuracy for forms with even (resp. odd) parity. We apply the trained LDA model to forms with unknown Fricke signs to make predictions. The average values based on the predicted Fricke signs are computed and compared to those for forms with known signs to verify the reasonableness of the predictions. Additionally, a subset of these predictions is evaluated against heuristic guesses provided by Hejhal's algorithm, showing a match approximately 95% of the time. We also use neural networks to obtain results comparable to those from the LDA model.

ORACLE: A Real-Time, Hierarchical, Deep-Learning Photometric Classifier for the LSST
Ved G. Shah, Alex Gagliano, Konstantin Malanchev, Gautham Narayan, The LSST Dark Energy Science Collaboration
[ arXiv:2501.01496 | code ]
Astrophysics

Abstract We present ORACLE, the first hierarchical deep-learning model for real-time, context-aware classification of transient and variable astrophysical phenomena. ORACLE is a recurrent neural network with Gated Recurrent Units (GRUs), and has been trained using a custom hierarchical cross-entropy loss function to provide high-confidence classifications along an observationally-driven taxonomy with as little as a single photometric observation. Contextual information for each object, including host galaxy photometric redshift, offset, ellipticity and brightness, is concatenated to the light curve embedding and used to make a final prediction. Training on ∼0.5M events from the Extended LSST Astronomical Time-Series Classification Challenge, we achieve a top-level (Transient vs Variable) macro-averaged precision of 0.96 using only 1 day of photometric observations after the first detection in addition to contextual information, for each event; this increases to >0.99 once 64 days of the light curve has been obtained, and 0.83 at 1024 days after first detection for 19-way classification (including supernova sub-types, active galactic nuclei, variable stars, microlensing events, and kilonovae). We also compare ORACLE with other state-of-the-art classifiers and report comparable performance for the 19-way classification task, in addition to delivering accurate top-level classifications much earlier.

A Near-IR Search for Helium in the Superluminous Supernova SN 2024ahr
Harsh Kumar, Edo Berger, Peter K. Blanchard, Sebastian Gomez, Daichi Hiramatsu, Moira Andrews, K. Azalee Bostroem, Yize Dong, Joseph Farah, Estefania Padilla Gonzalez, D. Andrew Howell, Curtis McCully, Darshana Mehta, Megan Newsome, Aravind P. Ravi, Giacomo Terreran
[ arXiv:2501.01485 ]
Astrophysics

Abstract We present a detailed study of SN 2024ahr, a hydrogen-poor superluminous supernova (SLSN-I), for which we determine a redshift of z=0.0861. SN 2024ahr has a peak absolute magnitude of Mg≈Mr≈−21 mag, rest-frame rise and decline times (50% of peak) of about 40 and 80 days, respectively, and typical spectroscopic evolution in the optical band. Similarly, modeling of the UV/optical light curves with a magnetar spin-down engine leads to typical parameters: an initial spin period of ≈3.3 ms, a magnetic field strength of ≈6×1013 G, and an ejecta mass of ≈9.5 M⊙. Due to its relatively low redshift we obtained a high signal-to-noise ratio near-IR spectrum about 43 rest-frame days post-peak to search for the presence of helium. We do not detect any significant feature at the location of the He I λ2.058 μm feature, and place a conservative upper limit of ∼0.05 M⊙ on the mass of helium in the outer ejecta. We detect broad features of Mg I λ1.575 μm and a blend of Co II λ2.126 μm and Mg II, λ2.136 μm, which are typical of Type Ic SNe, but with higher velocities. Examining the sample of SLSNe-I with NIR spectroscopy, we find that, unlike SN 2024ahr, these events are generally peculiar. This highlights the need for a large sample of prototypical SLSNe-I with NIR spectroscopy to constrain the fraction of progenitors with helium (Ib-like) and without helium (Ic-like) at the time of the explosion, and hence the evolutionary path(s) leading to the rare outcome of SLSNe-I.

Bootstrapped Reward Shaping
Jacob Adamczyk, Volodymyr Makarenko, Stas Tiomkin, Rahul V. Kulkarni
Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39, No. 15, 2025 [ arXiv:2501.00989 | code ] published
Foundational AI

Abstract In reinforcement learning, especially in sparse-reward domains, many environment steps are required to observe reward information. In order to increase the frequency of such observations, 'potential-based reward shaping' (PBRS) has been proposed as a method of providing a more dense reward signal while leaving the optimal policy invariant. However, the required 'potential function' must be carefully designed with task-dependent knowledge to not deter training performance. In this work, we propose a 'bootstrapped' method of reward shaping, termed BSRS, in which the agent's current estimate of the state-value function acts as the potential function for PBRS. We provide convergence proofs for the tabular setting, give insights into training dynamics for deep RL, and show that the proposed method improves training speed in the Atari suite.

Data for Mathematical Copilots: Better Ways of Presenting Proofs for Machine Learning
Simon Frieder, Jonas Bayer, Katherine M. Collins, Julius Berner, Jacob Loader, András Juhász, Fabian Ruehle, Sean Welleck, Gabriel Poesia, Ryan-Rhys Griffiths, Adrian Weller, Anirudh Goyal, Thomas Lukasiewicz, Timothy Gowers
[ arXiv:2412.15184 ]
Foundational AI

Abstract The suite of datasets commonly used to train and evaluate the mathematical capabilities of AI-based mathematical copilots (primarily large language models) exhibit several shortcomings. These limitations include a restricted scope of mathematical complexity, typically not exceeding lower undergraduate-level mathematics, binary rating protocols and other issues, which makes comprehensive proof-based evaluation suites difficult. We systematically explore these limitations and contend that enhancing the capabilities of large language models, or any forthcoming advancements in AI-based mathematical assistants (copilots or 'thought partners'), necessitates a paradigm shift in the design of mathematical datasets and the evaluation criteria of mathematical ability: It is necessary to move away from result-based datasets (theorem statement to theorem proof) and convert the rich facets of mathematical research practice to data LLMs can train on. Examples of these are mathematical workflows (sequences of atomic, potentially subfield-dependent tasks that are often performed when creating new mathematics), which are an important part of the proof-discovery process. Additionally, we advocate for mathematical dataset developers to consider the concept of 'motivated proof', introduced by G. Pólya in 1949, which can serve as a blueprint for datasets that offer a better proof learning signal, alleviating some of the mentioned limitations. Lastly, we introduce math datasheets for datasets, extending the general, dataset-agnostic variants of datasheets: We provide a questionnaire designed specifically for math datasets that we urge dataset creators to include with their datasets. This will make creators aware of potential limitations of their datasets while at the same time making it easy for readers to assess it from the point of view of training and evaluating mathematical copilots.

Constraints and Sensitivities for Dipole-Portal Heavy Neutral Leptons from ND280 and its Upgrade
Ming-Shau Liu, Nicholas Kamp, Carlos A. Argüelles
[ arXiv:2412.15051 ]
Experimental Physics

Abstract We report new constraints and sensitivities to heavy neutral leptons (HNLs) with transition magnetic moments, also known as dipole-portal HNLs. This is accomplished using data from the T2K ND280 near detector in addition to the projected three-year dataset of the upgraded ND280 detector. Dipole-portal HNLs have been extensively studied in the literature and offer a potential explanation for the 4.8σ MiniBooNE anomaly. To perform our analysis, we simulate HNL decays to e+e− pairs in the gaseous time projection chambers of the ND280 detector and its upgrade. Recasting an ND280 search for mass-mixed HNLs, we find that ND280 data places world-leading constraints on dipole-portal HNLs in the 390-743 MeV mass range, disfavoring the region of parameter space favored by the MiniBooNE anomaly. The addition of three years of ND280 upgrade data will be able to disfavor the MiniBooNE solution at the 5σ confidence level and extend the world-leading constraints to dipole-portal HNLs in the 148-860 MeV mass range. Our analysis suggests that ND280 data excludes dipole-portal HNLs as a solution to the MiniBooNE excess, motivating a dedicated search within the T2K collaboration and potentially highlighting the need for alternative explanations for the MiniBooNE anomaly.

The Neutrino Slice at Muon Colliders
Luc Bojorquez-Lopez, Matheus Hostert, Carlos A. Argüelles, Zhen Liu
[ arXiv:2412.14115 | code ]
Experimental Physics

Abstract Muon colliders provide an exciting new path pushing forward the energy frontier of particle physics. We point out a new use of these facilities for neutrino physics and beyond the Standard Model physics using their main detectors. Muon decays along the main accelerator rings induce an intense, highly collimated beam of neutrinos. As this beam crosses a thin slice of the kt-scale detector, it would induce unprecedented numbers of neutrino interactions, with O(104) events per second for a 10 TeV µ+µ− collider. We characterize these events, showing that they are highly energetic and possess a distinct timing signature with a large transverse displacement. We discuss promising applications of these events for instrumentation, electroweak, and beyond-the-Standard Model physics. For instance, we show that a sub-percent measurement of the neutrino-electron scattering rate enables new precision measurements of the Weak angle and a novel detection of the neutrino charge radius.

The Impact of Host-galaxy Properties on Supernova Classification with Hierarchical Labels
Villar, V. Ashley, Gomes, Sebastian, Berger, Edo, Gagliano,Alex
The Astrophysical Journal Supplement Series, Voulume 276, Number 1, 2024 [ ] published
Astrophysics

Abstract With the advent of the Vera C. Rubin Observatory, the discovery rate of supernovae (SNe) will surpass the rate of SNe with real time spectroscopic follow-up by 3 orders of magnitude. Accurate photometric classifiers are essential to both select interesting events for follow-up in real time and for archival population-level studies. In this work, we investigate the impact of observable host-galaxy information on the classification of SNe, both with and without additional light-curve and redshift information. We find that host-galaxy information alone can successfully isolate relatively pure (>90%) samples of Type Ia SNe with or without redshift information. With redshift information, we can additionally produce somewhat pure (>70%) samples of Type II SNe and superluminous SNe. Additionally with redshift information, host-galaxy properties do not significantly improve the accuracy of SN classification when paired with complete light curves. In the absence of redshift information, however, galaxy properties significantly increase the accuracy of photometric classification. As a part of this analysis, we present the first formal application of a new objective function, the weighted hierarchical cross entropy, to the problem of SN classification. This objective function more naturally accounts for the hierarchical nature of SN classes and, more broadly, transients. Finally, we present a new set of SN classifications for the Pan-STARRS Medium Deep Survey of SNe that lack spectroscopic redshift, increasing the full photometric sample to >4400 events.

Product Manifold Machine Learning for Physics
Nathaniel S. Woodward, Sang Eon Park, Gaia Grosso, Jeffrey Krupa, Philip Harris
[ arXiv:2412.07033 ]
Experimental Physics

Abstract Physical data are representations of the fundamental laws governing the Universe, hiding complex compositional structures often well captured by hierarchical graphs. Hyperbolic spaces are endowed with a non-Euclidean geometry that naturally embeds those structures. To leverage the benefits of non-Euclidean geometries in representing natural data we develop machine learning on PM spaces, Cartesian products of constant curvature Riemannian manifolds. As a use case we consider the classification of “jets”, sprays of hadrons and other subatomic particles produced by the hadronization of quarks and gluons in collider experiments. We compare the performance of PM-MLP and PM-Transformer models across several possible representations. Our experiments show that PM representations generally perform equal or better to fully Euclidean models of similar size, with the most significant gains found for highly hierarchical jets and small models. We discover significant correlation between the degree of hierarchical structure at a per-jet level and classification performance with the PM-Transformer in top tagging benchmarks. This is a promising result highlighting a potential direction for further improving machine learning model performance through tailoring geometric representation at a per-sample level in hierarchical datasets. These results reinforce the view of geometric representation as a key parameter in maximizing both performance and efficiency of machine learning on natural data.

Grayscale to Hyperspectral at Any Resolution Using a Phase-Only Lens
Dean Hazineh, Federico Capasso, Todd Zickler
[ arXiv:2412.02798 ]
Foundational AI

Abstract We consider the problem of reconstructing a H×W×31 hyperspectral image from a H×W grayscale snapshot measurement that is captured using a single diffractive optic and a filterless panchromatic photosensor. This problem is severely ill-posed, and we present the first model that is able to produce high-quality results. We train a conditional denoising diffusion model that maps a small grayscale measurement patch to a hyperspectral patch. We then deploy the model to many patches in parallel, using global physics-based guidance to synchronize the patch predictions. Our model can be trained using small hyperspectral datasets and then deployed to reconstruct hyperspectral images of arbitrary size. Also, by drawing multiple samples with different seeds, our model produces useful uncertainty maps. We show that our model achieves state-of-the-art performance on previous snapshot hyperspectral benchmarks where reconstruction is better conditioned. Our work lays the foundation for a new class of high-resolution hyperspectral imagers that are compact and light-efficient.

The Millennium and Astrid galaxies in effective field theory: comparison with galaxy-halo connection models at the field level
Mikhail M. Ivanov, Carolina Cuesta-Lazaro, Andrej Obuljen, Michael W. Toomey, Yueying Ni, Sownak Bose, Boryana Hadzhiyska, César Hernández-Aguayo, Lars Hernquist, Rahul Kannan, Volker Springel
[ arXiv:2412.01888 ]
Astrophysics

Abstract Cosmological analyses of redshift space clustering data are primarily based on using luminous ``red'' galaxies (LRGs) and ``blue'' emission line galaxies (ELGs) to trace underlying dark matter. Using the large high-fidelity high-resolution MillenniumTNG (MTNG) and Astrid simulations, we study these galaxies with the effective field theory (EFT)-based field level forward model. We confirm that both red and blue galaxies can be accurately modeled with EFT at the field level and their parameters match those of the phenomenological halo-based models. Specifically, we consider the state of the art Halo Occupation Distribution (HOD) and High Mass Quenched (HMQ) models for the red and blue galaxies, respectively. Our results explicitly confirm the validity of the halo-based models on large scales beyond the two-point statistics. In addition, we validate the field-level HOD/HMQ-based priors for EFT full-shape analysis. We find that the local bias parameters of the ELGs are in tension with the predictions of the LRG-like HOD models and present a simple analytic argument explaining this phenomenology. We also confirm that ELGs exhibit weaker non-linear redshift-space distortions (``fingers-of-God''), suggesting that a significant fraction of their data should be perturbative. We find that the response of EFT parameters to galaxy selection is sensitive to assumptions about baryonic feedback, suggesting that a detailed understanding of feedback processes is necessary for robust predictions of EFT parameters. Finally, using neural density estimation based on paired HOD-EFT parameter samples, we obtain optimal HOD models that reproduce the clustering of Astrid and MTNG galaxies.

Probing primordial non-Gaussianity by reconstructing the initial conditions
Xinyi Chen, Nikhil Padmanabhan, Daniel J. Eisenstein
[ arXiv:2412.00968 ]
Astrophysics

Abstract We propose to constrain the primordial (local-type) non-Gaussianity signal by first reconstructing the initial density field to remove the late time non-Gaussianities introduced by gravitational evolution. Our reconstruction algorithm combines perturbation theory on large scales with a convolutional neural network on small scales. We reconstruct the squared potential (that sources the non-Gaussian signal) out to k=0.2 h/Mpc to an accuracy of 99.8%. We cross-correlate this squared potential field with the reconstructed density field and verify that this computationally inexpensive estimator has the same information content as the full matter bispectrum. As a proof of concept, our approach can yield up to a factor of three improvement in the fNL constraints, although it does not yet include the complications of galaxy bias or imperfections in the reconstruction. These potential improvements make it a promising alternative to current approaches to constraining primordial non-Gaussianity.

SymbolFit: Automatic Parametric Modeling with Symbolic Regression
Ho Fung Tsoi, Dylan Rankin, Cecile Caillol, Miles Cranmer, Sridhara Dasu, Javier Duarte, Philip Harris, Elliot Lipeles, Vladimir Loncar
[ arXiv:2411.09851 ]
Experimental Physics

Abstract We introduce SymbolFit, a framework that automates parametric modeling by using symbolic regression to perform a machine-search for functions that fit the data while simultaneously providing uncertainty estimates in a single run. Traditionally, constructing a parametric model to accurately describe binned data has been a manual and iterative process, requiring an adequate functional form to be determined before the fit can be performed. The main challenge arises when the appropriate functional forms cannot be derived from first principles, especially when there is no underlying true closed-form function for the distribution. In this work, we develop a framework that automates and streamlines the process by utilizing symbolic regression, a machine learning technique that explores a vast space of candidate functions without requiring a predefined functional form because the functional form itself is treated as a trainable parameter, making the process far more efficient and effortless than traditional regression methods. We demonstrate the framework in high-energy physics experiments at the CERN Large Hadron Collider (LHC) using five real proton-proton collision datasets from new physics searches, including background modeling in resonance searches for high-mass dijet, trijet, paired-dijet, diphoton, and dimuon events. We show that our framework can flexibly and efficiently generate a wide range of candidate functions that fit a nontrivial distribution well using a simple fit configuration that varies only by random seed, and that the same fit configuration, which defines a vast function space, can also be applied to distributions of different shapes, whereas achieving a comparable result with traditional methods would have required extensive manual effort.

Type IIn Supernovae. I. Uniform Light Curve Characterization and a Bimodality in the Radiated Energy Distribution
Daichi Hiramatsu, Edo Berger, Sebastian Gomez, Peter K. Blanchard, Harsh Kumar, Wasundara Athukoralalage
[ arXiv:2411.07287 ]
Astrophysics

Abstract We present the largest uniform study to date of Type IIn supernovae (SNe IIn), focusing in this first paper on the multi-band optical light curves of 487 SNe IIn. The sample, constructed from multiple surveys, extends to z≈0.8, with the majority of events at z≲0.3. We construct uniform multi-band and bolometric light curves using Gaussian process regression, and determine key observed properties in the rest-frame (e.g., peak luminosity, timescales, radiated energy). We find that SNe IIn span broad ranges in peak luminosity (∼1042−1044 erg s−1) and timescales (∼20−300 days above 50% of peak luminosity), but the sample divides into two clear groups in the luminosity-timescale phase-space around the median peak luminosity (≈1043 erg s−1): faint-fast and luminous-slow groups. This leads to a strong bimodality in the radiated energy distribution, with peaks at ∼1049 and ∼2×1050 erg, with the latter events having a characteristic timescale of ∼100 days, and the former appearing to bifurcate into two branches with timescales of ∼40 and ∼70 days. Therefore, SNe IIn exhibit at least two dominant groupings, and perhaps three, which are likely reflective of different progenitor and/or circumstellar medium formation pathways. We do not find any obvious transition in SN IIn properties at the arbitrary cut-off of ≈−20 mag used for the designation 'Type II Superluminous Supernovae', and we argue that this classification should be abandoned. The absence of SNe IIn with timescales of ≲14 days defines the region occupied by fast transients with evidence for interaction with hydrogen-poor circumstellar medium.

A Twist on Heterotic Little String Duality
Hamza Ahmed, Paul-Konstantin Oehlmann, Fabian Ruehle
[ arXiv:2411.05313 ]
Theoretical Physics

Abstract In this work, we significantly expand the web of T-dualities among heterotic NS5-brane theories with eight supercharges. This is achieved by introducing twists involving outer automorphisms of discrete gauge/flavor factors and tensor multiplet permutations along the compactification circle. We assemble field theory data that we propose as invariants across T-dual theories, comprised of twisted Coulomb branch dimensions, higher group structures and flavor symmetry ranks. Using this data, we establish a detailed field theory correspondence between singularities of the compactification space, the number five-branes in the theory, and the flavor symmetry factors. The twisted theories are realized via M-theory compactifications on non-compact genus-one fibered Calabi-Yau threefolds without section. This approach allows us to prove duality of twisted and (un-)twisted theories by leveraging M/F-theory duality and identifying inequivalent torus fibrations in the same geometry. We construct several new 5D theories, including a novel type of CHL-like twisted theory where the two M9 branes are identified. Using their field theory invariants, we also construct their dual theories.

Conversations and Deliberations: Non-Standard Cosmological Epochs and Expansion Histories
Brian Batell, Keith R. Dienes, Brooks Thomas, Scott Watson, Rouzbeh Allahverdi, Mustafa Amin, Kimberly K. Boddy, M. Sten Delos, Adrienne L. Erickcek, Akshay Ghalsasi, John T. Giblin Jr., James Halverson, Fei Huang, Andrew J. Long, Lauren Pearce, Barmak Shams Es Haghi, Jessie Shelton, Gary Shiu, Kuver Sinha, Tristan L. Smith
[ arXiv:2411.04780 ]
Astrophysics

Abstract This document summarizes the discussions which took place during the PITT-PACC Workshop entitled 'Non-Standard Cosmological Epochs and Expansion Histories,' held in Pittsburgh, Pennsylvania, Sept. 5-7, 2024. Much like the non-standard cosmological epochs that were the subject of these discussions, the format of this workshop was also non-standard. Rather than consisting of a series of talks from participants, with each person presenting their own work, this workshop was instead organized around free-form discussion blocks, with each centered on a different overall theme and guided by a different set of Discussion Leaders. This document is not intended to serve as a comprehensive review of these topics, but rather as an informal record of the discussions that took place during the workshop, in the hope that the content and free-flowing spirit of these discussions may inspire new ideas and research directions.

Adaptive Length Image Tokenization via Recurrent Allocation
Shivam Duggal, Phillip Isola, Antonio Torralba, William T. Freeman
ICLR 2025 [ arXiv:2411.02393 | code ] published
Foundational AI

Abstract Current vision systems typically assign fixed-length representations to images, regardless of the information content. This contrasts with human intelligence - and even large language models - which allocate varying representational capacities based on entropy, context and familiarity. Inspired by this, we propose an approach to learn variable-length token representations for 2D images. Our encoder-decoder architecture recursively processes 2D image tokens, distilling them into 1D latent tokens over multiple iterations of recurrent rollouts. Each iteration refines the 2D tokens, updates the existing 1D latent tokens, and adaptively increases representational capacity by adding new tokens. This enables compression of images into a variable number of tokens, ranging from 32 to 256. We validate our tokenizer using reconstruction loss and FID metrics, demonstrating that token count aligns with image entropy, familiarity and downstream task requirements. Recurrent token processing with increasing representational capacity in each iteration shows signs of token specialization, revealing potential for object / part discovery.

DexHub and DART: Towards Internet Scale Robot Data Collection
Younghyo Park, Jagdeep Singh Bhatia, Lars Ankile, Pulkit Agrawal
[ arXiv:2411.02214 ]
Foundational AI

Abstract The quest to build a generalist robotic system is impeded by the scarcity of diverse and high-quality data. While real-world data collection effort exist, requirements for robot hardware, physical environment setups, and frequent resets significantly impede the scalability needed for modern learning frameworks. We introduce DART, a teleoperation platform designed for crowdsourcing that reimagines robotic data collection by leveraging cloud-based simulation and augmented reality (AR) to address many limitations of prior data collection efforts. Our user studies highlight that DART enables higher data collection throughput and lower physical fatigue compared to real-world teleoperation. We also demonstrate that policies trained using DART-collected datasets successfully transfer to reality and are robust to unseen visual disturbances. All data collected through DART is automatically stored in our cloud-hosted database, DexHub, which will be made publicly available upon curation, paving the path for DexHub to become an ever-growing data hub for robot learning. Videos are available at: https://dexhub.ai/project

Small-scale Hamiltonian optimization of interpolating operators for Lagrangian lattice quantum field theory
Artur Avkhadiev, Lena Funcke, Karl Jansen, Stefan Kühn, Phiala E. Shanahan
[ arXiv:2411.02185 ]
Theoretical Physics

Abstract Lattice quantum field theory calculations may potentially combine the advantages of Hamiltonian formulations with the scalability and control of conventional Lagrangian frameworks. However, such hybrid approaches need to consider (1) the differences in renormalized coupling values between the two formulations, and (2) finite-volume and discretization effects when the Hamiltonian component of the calculation is characterized by a smaller volume or coarser lattice spacing than the Lagrangian component. This work investigates the role of both factors in the application of Hamiltonian-optimized interpolating operator constructions for the conventional Lagrangian framework. The numerical investigation is realized for the pseudoscalar meson in the Schwinger model, using tensor-network and Monte-Carlo calculations. It is demonstrated that tensor-network-optimized constructions are robust to both (1) and (2). In particular, accurate optimized constructions for the pseudoscalar meson can be obtained from calculations with a smaller number of Hamiltonian lattice sites, even when the meson mass itself receives significant finite-volume corrections. To the extent that these results generalize to theories with more complicated spectra, the method holds promise for near-term applications in large-scale calculations of lattice quantum field theory.

Not So Flat Metrics
Cristofero S. Fraser-Taliente, Thomas R. Harvey, Manki Kim
[ arXiv:2411.00962 ]
Theoretical Physics

Abstract In order to be in control of the α′ derivative expansion, geometric string compactifications are understood in the context of a large volume approximation. In this letter, we consider the reduction of these higher derivative terms, and propose an improved estimate on the large volume approximation using numerical Calabi-Yau metrics obtained via machine learning methods. Further to this, we consider the α′3 corrections to numerical Calabi-Yau metrics in the context of IIB string theory. This correction represents one of several important contributions for realistic string compactifications -- alongside, for example, the backreaction of fluxes and local sources -- all of which have important consequences for string phenomenology. As a simple application of the corrected metric, we compute the change to the spectrum of the scalar Laplacian.

A Lorentz-Equivariant Transformer for All of the LHC
Johann Brehmer, Víctor Bresó, Pim de Haan, Tilman Plehn, Huilin Qu, Jonas Spinner, Jesse Thaler
[ arXiv:2411.00446 | code ]
Foundational AI

Abstract We show that the Lorentz-Equivariant Geometric Algebra Transformer (L-GATr) yields state-of-the-art performance for a wide range of machine learning tasks at the Large Hadron Collider. L-GATr represents data in a geometric algebra over space-time and is equivariant under Lorentz transformations. The underlying architecture is a versatile and scalable transformer, which is able to break symmetries if needed. We demonstrate the power of L-GATr for amplitude regression and jet classification, and then benchmark it as the first Lorentz-equivariant generative network. For all three LHC tasks, we find significant improvements over previous architectures.

Linearized Wasserstein Barycenters: Synthesis, Analysis, Representational Capacity, and Applications
Matthew Werenski, Brendan Mallery, Shuchin Aeron, James M. Murphy
[ arXiv:2410.23602 | code ]
Foundational AI

Abstract We propose the extit{linear barycentric coding model (LBCM)} that utilizes the linear optimal transport (LOT) metric for analysis and synthesis of probability measures. We provide a closed-form solution to the variational problem characterizing the probability measures in the LBCM and establish equivalence of the LBCM to the set of Wasserstein-2 barycenters in the special case of compatible measures. Computational methods for synthesizing and analyzing measures in the LBCM are developed with finite sample guarantees. One of our main theoretical contributions is to identify an LBCM, expressed in terms of a simple family, which is sufficient to express all probability measures on the interval [0,1]. We show that a natural analogous construction of an LBCM in ℝ2 fails, and we leave it as an open problem to identify the proper extension in more than one dimension. We conclude by demonstrating the utility of LBCM for covariance estimation and data imputation.

Inferring the Morphology of the Galactic Center Excess with Gaussian Processes
Edward D. Ramirez, Yitian Sun, Matthew R. Buckley, Siddharth Mishra-Sharma, Tracy R. Slatyer
Physical Review D, 2025, Vol. 111, Iss. 6 [ arXiv:2410.21367 | code ] published
Astrophysics

Abstract Descriptions of the Galactic Center using Fermi gamma-ray data have so far modeled the Galactic Center Excess (GCE) as a template with fixed spatial morphology or as a linear combination of such templates. Although these templates are informed by various physical expectations, the morphology of the excess is a priori unknown. For the first time, we describe the GCE using a flexible, non-parametric machine learning model -- the Gaussian process (GP). We assess our model's performance on synthetic data, demonstrating that the model can recover the templates used to generate the data. We then fit the \Fermi data with our model in a single energy bin from 2-20 GeV (leaving a spectral GP analysis of the GCE for future work) using a variety of template models of diffuse gamma-ray emission to quantify our fits' systematic uncertainties associated with diffuse emission modeling. We interpret our best-fit GP in terms of GCE templates consisting of an NFW squared template and a bulge component to determine which bulge models can best describe the fitted GP and to what extent the best-fit GP is described better by an NFW squared template versus a bulge template. The best-fit GP contains morphological features that are typically not associated with traditional GCE studies. These include a localized bright source at around (ℓ,b)=(20∘,0∘) and a diagonal arm extending Northwest from the Galactic Center. In spite of these novel features, the fitted GP is explained best by a template-based model consisting of the bulge presented in Coleman et al. (2020) and a squared NFW component. Our results suggest that the physical interpretation of the GCE in terms of stellar bulge and NFW-like components is highly sensitive to the assumed morphologies, background models, and the region of the sky used for inference.

A Cosmic-Scale Benchmark for Symmetry-Preserving Data Processing
Julia Balla, Siddharth Mishra-Sharma, Carolina Cuesta-Lazaro, Tommi Jaakkola, Tess Smidt
[ arXiv:2410.20516 | code ]
Foundational AI

Abstract Efficiently processing structured point cloud data while preserving multiscale information is a key challenge across domains, from graphics to atomistic modeling. Using a curated dataset of simulated galaxy positions and properties, represented as point clouds, we benchmark the ability of graph neural networks to simultaneously capture local clustering environments and long-range correlations. Given the homogeneous and isotropic nature of the Universe, the data exhibits a high degree of symmetry. We therefore focus on evaluating the performance of Euclidean symmetry-preserving (E(3)-equivariant) graph neural networks, showing that they can outperform non-equivariant counterparts and domain-specific information extraction techniques in downstream performance as well as simulation-efficiency. However, we find that current architectures fail to capture information from long-range correlations as effectively as domain-specific baselines, motivating future work on architectures better suited for extracting long-range information.

Joint Modeling of Quasar Variability and Accretion Disk Reprocessing using Latent Stochastic Differential Equation
Joshua Fagin, James Hung-Hsu Chan, Henry Best, Matthew O’Dowd, K. E. Saavik Ford, Matthew J. Graham, Ji Won Park, V. Ashley Villar
[ arXiv:2410.18423 | code ]
Astrophysics

Abstract Quasars are bright active galactic nuclei powered by the accretion of matter around supermassive black holes at the center of galaxies. Their stochastic brightness variability depends on the physical properties of the accretion disk and black hole. The upcoming Rubin Observatory Legacy Survey of Space and Time (LSST) is expected to observe tens of millions of quasars, so there is a need for efficient techniques like machine learning that can handle the large volume of data. Quasar variability is believed to be driven by an X-ray corona, which is reprocessed by the accretion disk and emitted as UV/optical variability. We are the first to introduce an auto-differentiable simulation of the accretion disk and reprocessing. We use the simulation as a direct component of our neural network to jointly model the driving variability and reprocessing to fit simulated LSST 10-year quasar light curves. The driving variability is reconstructed using a latent stochastic differential equation, a physically motivated, generative deep learning method that can model continuous-time stochastic dynamics. By embedding these physical processes into our network, we achieve a model that is more robust and interpretable. We also use transformers to scale our model to tens of millions of parameters. We demonstrate how our model outperforms a Gaussian process regression baseline and can infer accretion disk parameters and time delays between wavebands, even for out-of-distribution driving signals. Our approach provides a powerful and scalable framework that can be adapted to solve other inverse problems in multivariate time series with irregular sampling.

Fermion Masses and Mixing in String-Inspired Models
Andrei Constantin, Cristofero S. Fraser-Taliente, Thomas R. Harvey, Lucas T. Y. Leung, Andre Lukas
[ arXiv:2410.17704 ]
Theoretical Physics

Abstract We study a class of supersymmetric Froggatt-Nielsen (FN) models with multiple U(1) symmetries and Standard Model (SM) singlets inspired by heterotic string compactifications on Calabi-Yau threefolds. The string-theoretic origin imposes a particular charge pattern on the SM fields and FN singlets, dividing the latter into perturbative and non-perturbative types. Employing systematic and heuristic search strategies, such as genetic algorithms, we identify charge assignments and singlet VEVs that replicate the observed mass and mixing hierarchies in the quark sector, and subsequently refine the Yukawa matrix coefficients to accurately match the observed values for the Higgs VEV, the quark and charged lepton masses and the CKM matrix. This bottom-up approach complements top-down string constructions and our results demonstrate that string FN models possess a sufficiently rich structure to account for flavour physics. On the other hand, the limited number of distinct viable charge patterns identified here indicates that flavour physics imposes tight constraints on string theory models, adding new constraints on particle spectra that are essential for achieving a realistic phenomenology.

Blast: a Web Application for Characterizing the Host Galaxies of Astrophysical Transients
D. O. Jones, P. McGill, T. A. Manning, A. Gagliano, B. Wang, D. A. Coulter, R. J. Foley, G. Narayan, V. A. Villar, L. Braff, A. W. Engel, D. Farias, Z. Lai, K. Loertscher, J. Kutcka, S. Thorp, J. Vazquez
[ arXiv:2410.17322 ]
Astrophysics

Abstract Characterizing the host galaxies of astrophysical transients is important to many areas of astrophysics, including constraining the progenitor systems of core-collapse supernovae, correcting Type Ia supernova distances, and probabilistically classifying transients without photometric or spectroscopic data. Given the increasing transient discovery rate in the coming years, there is substantial utility in providing public, transparent, reproducible, and automatic characterization for large samples of transient host galaxies. Here we present Blast, a web application that ingests live streams of transient alerts, matches transients to their host galaxies, and performs photometry on coincident archival imaging data of the host galaxy. The photometry is then used to infer both global host-galaxy properties and galaxy properties within 2 kpc of the transient location by using the Prospector Bayesian inference framework, with an acceleration in evaluation speed achieved via simulation-based inference. Blast provides host-galaxy properties to users via a web browser or an application program interface. The software can be extended to support alternative photometric or SED-fitting algorithms, and can be scaled via an asynchronous worker queue across multiple compute nodes to handle the processing of large volumes of transient alerts for upcoming transient surveys. Blast has been ingesting newly discovered transients from the Transient Name Server since mid-2024, and has currently measured SED parameters for more than 6000 transients. The service is publicly available at this https URL: https://blast.scimma.org/.

Decomposing The Dark Matter of Sparse Autoencoders
Joshua Engels, Logan Riggs, Max Tegmark
[ arXiv:2410.14670 | code ]
Foundational AI

Abstract Sparse autoencoders (SAEs) are a promising technique for decomposing language model activations into interpretable linear features. However, current SAEs fall short of completely explaining model performance, resulting in 'dark matter': unexplained variance in activations. This work investigates dark matter as an object of study in its own right. Surprisingly, we find that much of SAE dark matter--about half of the error vector itself and >90% of its norm--can be linearly predicted from the initial activation vector. Additionally, we find that the scaling behavior of SAE error norms at a per token level is remarkably predictable: larger SAEs mostly struggle to reconstruct the same contexts as smaller SAEs. We build on the linear representation hypothesis to propose models of activations that might lead to these observations, including postulating a new type of 'introduced error'; these insights imply that the part of the SAE error vector that cannot be linearly predicted ('nonlinear' error) might be fundamentally different from the linearly predictable component. To validate this hypothesis, we empirically analyze nonlinear SAE error and show that 1) it contains fewer not yet learned features, 2) SAEs trained on it are quantitatively worse, 3) it helps predict SAE per-token scaling behavior, and 4) it is responsible for a proportional amount of the downstream increase in cross entropy loss when SAE activations are inserted into the model. Finally, we examine two methods to reduce nonlinear SAE error at a fixed sparsity: inference time gradient pursuit, which leads to a very slight decrease in nonlinear error, and linear transformations from earlier layer SAE outputs, which leads to a larger reduction.

Learning Efficient Representations of Neutrino Telescope Events
Felix J. Yu, Nicholas Kamp, Carlos A. Argüelles
[ arXiv:2410.13148 | code ]
Experimental Physics

Abstract Neutrino telescopes detect rare interactions of particles produced in some of the most extreme environments in the Universe. This is accomplished by instrumenting a cubic-kilometer volume of naturally occurring transparent medium with light sensors. Given their substantial size and the high frequency of background interactions, these telescopes amass an enormous quantity of large variance, high-dimensional data. These attributes create substantial challenges for analyzing and reconstructing interactions, particularly when utilizing machine learning (ML) techniques. In this paper, we present a novel approach, called om2vec, that employs transformer-based variational autoencoders to efficiently represent neutrino telescope events by learning compact and descriptive latent representations. We demonstrate that these latent representations offer enhanced flexibility and improved computational efficiency, thereby facilitating downstream tasks in data analysis.

A Hodge Theory for Entanglement Cohomology
Christian Ferko, Eashan Iyer, Kasra Mossayebi, Gregor Sanfey
Physical Review A, 2025, Volume 111, Issue 3 [ arXiv:2410.12529 ] published
Theoretical Physics

Abstract We explore and extend the application of homological algebra to describe quantum entanglement, initiated in arXiv:1901.02011, focusing on the Hodge-theoretic structure of entanglement cohomology in finite-dimensional quantum systems. We construct analogues of the Hodge star operator, inner product, codifferential, and Laplacian for entanglement k-forms. We also prove that such k-forms obey versions of the Hodge isomorphism theorem and Hodge decomposition, and that they exhibit Hodge duality. As a corollary, we conclude that the dimensions of the k-th and (n−k)-th cohomologies coincide for entanglement in n-partite pure states, which explains a symmetry property ('Poincare duality') of the associated Poincare polynomials.

Auriga Streams II: orbital properties of tidally disrupting satellites of Milky Way-mass galaxies
Nora Shipp, Alexander H. Riley, Christine M. Simpson, Rebekka Bieri, Lina Necib, Arpit Arora, Francesca Fragkoudi, Facundo A. Gómez, Robert J. J. Grand, Federico Marinacci
[ arXiv:2410.09143 ]
Astrophysics

Abstract Galaxies like the Milky Way are surrounded by complex populations of satellites at all stages of tidal disruption. In this paper, we present a dynamical study of the disrupting satellite galaxies in the Auriga simulations that are orbiting 28 distinct Milky Way-mass hosts across three resolutions. We find that the satellite galaxy populations are highly disrupted. The majority of satellites that remain fully intact at present day were accreted recently without experiencing more than one pericentre (nperi≲1) and have large apocentres (rapo≳200 kpc) and pericentres (rperi≳50 kpc). The remaining satellites have experienced significant tidal disruption and, given full knowledge of the system, would be classified as stellar streams. We find stellar streams in Auriga across the range of pericentres and apocentres of the known Milky Way dwarf galaxy streams and, interestingly, overlapping significantly with the Milky Way intact satellite population. We find no significant change in satellite orbital distributions across resolution. However, we do see substantial halo-to-halo variance of (rperi,rapo) distributions across host galaxies, as well as a dependence of satellite orbits on host halo mass - systems disrupt at larger pericentres and apocentres in more massive hosts. Our results suggest that either cosmological simulations (including, but not limited to, Auriga) are disrupting satellites far too readily, or that the Milky Way's satellites are more disrupted than current imaging surveys have revealed. Future observing facilities and careful mock observations of these systems will be key to revealing the nature of this apparent discrepancy.

Reinforcement Learning for Control of Non-Markovian Cellular Population Dynamics
Josiah C. Kratz, Jacob Adamczyk
[ arXiv:2410.08439 | code ]
Foundational AI

Abstract Many organisms and cell types, from bacteria to cancer cells, exhibit a remarkable ability to adapt to fluctuating environments. Additionally, cells can leverage memory of past environments to better survive previously-encountered stressors. From a control perspective, this adaptability poses significant challenges in driving cell populations toward extinction, and is thus an open question with great clinical significance. In this work, we focus on drug dosing in cell populations exhibiting phenotypic plasticity. For specific dynamical models switching between resistant and susceptible states, exact solutions are known. However, when the underlying system parameters are unknown, and for complex memory-based systems, obtaining the optimal solution is currently intractable. To address this challenge, we apply reinforcement learning (RL) to identify informed dosing strategies to control cell populations evolving under novel non-Markovian dynamics. We find that model-free deep RL is able to recover exact solutions and control cell populations even in the presence of long-range temporal dynamics.

Dirac Traces and the Tutte Polynomial
Joshua Lin
[ arXiv:2410.08161 ]
Theoretical Physics

Abstract Perturbative calculations involving fermion loops in quantum field theories require tracing over Dirac matrices. A simple way to regulate the divergences that generically appear in these calculations is dimensional regularisation, which has the consequence of replacing 4-dimensional Dirac matrices with d-dimensional counterparts for arbitrary complex values of d. In this work, a connection between traces of d-dimensional Dirac matrices and computations of the Tutte polynomial of associated graphs is proven. The time complexity of computing Dirac traces is analysed by this connection, and improvements to algorithms for computing Dirac traces are proposed.

Efficient Dictionary Learning with Switch Sparse Autoencoders
Anish Mudide, Joshua Engels, Eric J. Michaud, Max Tegmark, Christian Schroeder de Witt
ICLR, 2025 [ arXiv:2410.08201 | code ] published
Foundational AI

Abstract Sparse autoencoders (SAEs) are a recent technique for decomposing neural network activations into human-interpretable features. However, in order for SAEs to identify all features represented in frontier models, it will be necessary to scale them up to very high width, posing a computational challenge. In this work, we introduce Switch Sparse Autoencoders, a novel SAE architecture aimed at reducing the compute cost of training SAEs. Inspired by sparse mixture of experts models, Switch SAEs route activation vectors between smaller 'expert' SAEs, enabling SAEs to efficiently scale to many more features. We present experiments comparing Switch SAEs with other SAE architectures, and find that Switch SAEs deliver a substantial Pareto improvement in the reconstruction vs. sparsity frontier for a given fixed training compute budget. We also study the geometry of features across experts, analyze features duplicated across experts, and verify that Switch SAE features are as interpretable as features found by other SAE architectures.

Generalization from Starvation: Hints of Universality in LLM Knowledge Graph Learning
David D. Baek, Yuxiao Li, Max Tegmark
[ arXiv:2410.08255 ]
Foundational AI

Abstract Motivated by interpretability and reliability, we investigate how neural networks represent knowledge during graph learning, We find hints of universality, where equivalent representations are learned across a range of model sizes (from 102 to 109 parameters) and contexts (MLP toy models, LLM in-context learning and LLM training). We show that these attractor representations optimize generalization to unseen examples by exploiting properties of knowledge graph relations (e.g. symmetry and meta-transitivity). We find experimental support for such universality by showing that LLMs and simpler neural networks can be stitched, i.e., by stitching the first part of one model to the last part of another, mediated only by an affine or almost affine transformation. We hypothesize that this dynamic toward simplicity and generalization is driven by 'intelligence from starvation': where overfitting is minimized by pressure to minimize the use of resources that are either scarce or competed for against other tasks.

The Geometry of Concepts: Sparse Autoencoder Feature Structure
Yuxiao Li, Eric J. Michaud, David D. Baek, Joshua Engels, Xiaoqing Sun, Max Tegmark
Entropy, 2025, Volume 27, Issue 4 [ arXiv:2410.19750 ] published
Foundational AI

Abstract Sparse autoencoders have recently produced dictionaries of high-dimensional vectors corresponding to the universe of concepts represented by large language models. We find that this concept universe has interesting structure at three levels: 1) The 'atomic' small-scale structure contains 'crystals' whose faces are parallelograms or trapezoids, generalizing well-known examples such as (man-woman-king-queen). We find that the quality of such parallelograms and associated function vectors improves greatly when projecting out global distractor directions such as word length, which is efficiently done with linear discriminant analysis. 2) The 'brain' intermediate-scale structure has significant spatial modularity; for example, math and code features form a 'lobe' akin to functional lobes seen in neural fMRI images. We quantify the spatial locality of these lobes with multiple metrics and find that clusters of co-occurring features, at coarse enough scale, also cluster together spatially far more than one would expect if feature geometry were random. 3) The 'galaxy' scale large-scale structure of the feature point cloud is not isotropic, but instead has a power law of eigenvalues with steepest slope in middle layers. We also quantify how the clustering entropy depends on the layer.

RelitLRM: Generative Relightable Radiance for Large Reconstruction Models
Tianyuan Zhang, Zhengfei Kuang, Haian Jin, Zexiang Xu, Sai Bi, Hao Tan, He Zhang, Yiwei Hu, Milos Hasan, William T. Freeman, Kai Zhang, Fujun Luan
[ arXiv:2410.06231 ]
Foundational AI

Abstract We propose RelitLRM, a Large Reconstruction Model (LRM) for generating high-quality Gaussian splatting representations of 3D objects under novel illuminations from sparse (4-8) posed images captured under unknown static lighting. Unlike prior inverse rendering methods requiring dense captures and slow optimization, often causing artifacts like incorrect highlights or shadow baking, RelitLRM adopts a feed-forward transformer-based model with a novel combination of a geometry reconstructor and a relightable appearance generator based on diffusion. The model is trained end-to-end on synthetic multi-view renderings of objects under varying known illuminations. This architecture design enables to effectively decompose geometry and appearance, resolve the ambiguity between material and lighting, and capture the multi-modal distribution of shadows and specularity in the relit appearance. We show our sparse-view feed-forward RelitLRM offers competitive relighting results to state-of-the-art dense-view optimization-based baselines while being significantly faster. Our project page is available at: https://relit-lrm.github.io/

Transformers are Efficient Compilers, Provably
Xiyu Zhai, Runlong Zhou, Liao Zhang, Simon Shaolei Du
[ arXiv:2410.14706 ]
Foundational AI

Abstract Transformer-based large language models (LLMs) have demonstrated surprisingly robust performance across a wide range of language-related tasks, including programming language understanding and generation. In this paper, we take the first steps towards a formal investigation of using transformers as compilers from an expressive power perspective. To this end, we introduce a representative programming language, Mini-Husky, which encapsulates key features of modern C-like languages. We show that if the input code sequence has a bounded depth in both the Abstract Syntax Tree (AST) and type inference (reasonable assumptions based on the clean code principle), then the number of parameters required by transformers depends only on the logarithm of the input sequence length to handle compilation tasks, such as AST construction, symbol resolution, and type analysis. A significant technical challenge stems from the fact that transformers operate at a low level, where each layer processes the input sequence as raw vectors without explicitly associating them with predefined structure or meaning. In contrast, high-level compiler tasks necessitate managing intricate relationships and structured program information. Our primary technical contribution is the development of a domain-specific language, Cybertron, which generates formal proofs of the transformer's expressive power, scaling to address compiler tasks. We further establish that recurrent neural networks (RNNs) require at least a linear number of parameters relative to the input sequence, leading to an exponential separation between transformers and RNNs. Finally, we empirically validate our theoretical results by comparing transformers and RNNs on compiler tasks within Mini-Husky.

SPECTER: Efficient Evaluation of the Spectral EMD
Rikab Gambhir, Andrew J. Larkoski, Jesse Thaler
Journal of High Energy Physics, Volume 2024, article number 219, 2025 [ arXiv:2410.05379 | code ] published
Theoretical Physics

Abstract The Energy Mover's Distance (EMD) has seen use in collider physics as a metric between events and as a geometric method of defining infrared and collinear safe observables. Recently, the Spectral Energy Mover's Distance (SEMD) has been proposed as a more analytically tractable alternative to the EMD. In this work, we obtain a closed-form expression for the Riemannian-like p = 2 SEMD metric between events, eliminating the need to numerically solve an optimal transport problem. Additionally, we show how the SEMD can be used to define event and jet shape observables by minimizing the distance between events and parameterized energy flows (similar to the EMD), and we obtain closed-form expressions for several of these observables. We also present the SPECTER framework, an efficient and highly parallelized implementation of the SEMD metric and SEMD-derived shape observables as an analogue of the previously-introduced SHAPER for EMD-based computations. We demonstrate that computing the SEMD with SPECTER can be up to a thousand times faster than computing the EMD with standard optimal transport libraries.

Exploring gauge-fixing conditions with gradient-based optimization
William Detmold, Gurtej Kanwar, Yin Lin, Phiala E. Shanahan, Michael L. Wagmam
[ arXiv:2410.03602 ]
Theoretical Physics

Abstract Lattice gauge fixing is required to compute gauge-variant quantities, for example those used in RI-MOM renormalization schemes or as objects of comparison for model calculations. Recently, gauge-variant quantities have also been found to be more amenable to signal-to-noise optimization using contour deformations. These applications motivate systematic parameterization and exploration of gauge-fixing schemes. This work introduces a differentiable parameterization of gauge fixing which is broad enough to cover Landau gauge, Coulomb gauge, and maximal tree gauges. The adjoint state method allows gradient-based optimization to select gauge-fixing schemes that minimize an arbitrary target loss function.

Formation of Representations in Neural Networks
Liu Ziyin, Isaac Chuang, Tomer Galanti, Tomaso Poggio
[ arXiv:2410.03006 ]
Foundational AI

Abstract Understanding neural representations will help open the black box of neural networks and advance our scientific understanding of modern AI systems. However, how complex, structured, and transferable representations emerge in modern neural networks has remained a mystery. Building on previous results, we propose the Canonical Representation Hypothesis (CRH), which posits a set of six alignment relations to universally govern the formation of representations in most hidden layers of a neural network. Under the CRH, the latent representations (R), weights (W), and neuron gradients (G) become mutually aligned during training. This alignment implies that neural networks naturally learn compact representations, where neurons and weights are invariant to task-irrelevant transformations. We then show that the breaking of CRH leads to the emergence of reciprocal power-law relations between R, W, and G, which we refer to as the Polynomial Alignment Hypothesis (PAH). We present a minimal-assumption theory demonstrating that the balance between gradient noise and regularization is crucial for the emergence the canonical representation. The CRH and PAH lead to an exciting possibility of unifying major key deep learning phenomena, including neural collapse and the neural feature ansatz, in a single framework.

TT-like Flows of Yang-Mills Theories
Christian Ferko, Jue Hou, Tommaso Morone, Gabriele Tartaglino-Mazzucchelli, Roberto Tateo
Physical Review Letters, 2025, Vol.134, Iss.10 [ arXiv:2409.18740 ] published
Theoretical Physics

Abstract We study TT¯-like deformations of d>2 Yang-Mills theories. The standard TT¯ flows lead to multi-trace Lagrangians, and the non-Abelian gauge structures make it challenging to find Lagrangians in a closed form. However, within the geometric approach to TT¯, we obtain the closed-form solution to the metric flow and stress-energy tensor, and show that instanton solutions are undeformed. We also introduce new symmetrised single-trace TT¯-like deformations, whose solutions in d=4 include the non-Abelian Born-Infeld Lagrangian proposed by Tseytlin in 1997.

Optimal Quantum Purity Amplification
Zhaoyi Li, Honghao Fu, Takuya Isogawa, Isaac Chuang
[ arXiv:2409.18167 | code ]
Experimental Physics

Abstract Quantum purity amplification (QPA) offers a novel approach to counteracting the pervasive noise that degrades quantum states. We present the optimal QPA protocol for general quantum systems against global depolarizing noise, which has remained unknown for two decades. We construct and prove the optimality of our protocol, which demonstrates improved fidelity scaling compared to the best-known methods. We explore the operational interpretation of the protocol and provide simple examples of how to compile it into efficient circuits for near-term experiments. Furthermore, we conduct numerical simulations to investigate the effectiveness of our protocol in the quantum simulation of Hamiltonian evolution, demonstrating its ability to enhance fidelity even under circuit-level noise. Our findings suggest that QPA could improve the performance of quantum information processing tasks, particularly in the context of Noisy Intermediate-Scale Quantum (NISQ) devices, where reducing the effect of noise with limited resources is critical.

Seeing Faces in Things: A Model and Dataset for Pareidolia
Mark Hamilton, Simon Stent, Vasha DuTell, Anne Harrington, Jennifer Corbett, Ruth Rosenholtz, William T. Freeman
Springer Nature Link [ arXiv:2409.16143 | code ] published
Foundational AI

Abstract The human visual system is well-tuned to detect faces of all shapes and sizes. While this brings obvious survival advantages, such as a better chance of spotting unknown predators in the bush, it also leads to spurious face detections. ``Face pareidolia'' describes the perception of face-like structure among otherwise random stimuli: seeing faces in coffee stains or clouds in the sky. In this paper, we study face pareidolia from a computer vision perspective. We present an image dataset of ``Faces in Things'', consisting of five thousand web images with human-annotated pareidolic faces. Using this dataset, we examine the extent to which a state-of-the-art human face detector exhibits pareidolia, and find a significant behavioral gap between humans and machines. We find that the evolutionary need for humans to detect animal faces, as well as human faces, may explain some of this gap. Finally, we propose a simple statistical model of pareidolia in images. Through studies on human subjects and our pareidolic face detectors we confirm a key prediction of our model regarding what image conditions are most likely to induce pareidolia. Dataset and Website: this https URL

StreamGen: Connecting Populations of Streams and Shells to Their Host Galaxies
Adriana Dropulic, Nora Shipp, Stacy Kim, Zeineb Mezghanni, Lina Necib, Mariangela Lisanti
[ arXiv:2409.13810 | code ]
Astrophysics

Abstract In this work, we study how the abundance and dynamics of populations of disrupting satellite galaxies change systematically as a function of host galaxy properties. We apply a theoretical model of the phase-mixing process to classify intact satellite galaxies, stellar stream-like and shell-like debris in ~1500 Milky Way-mass systems generated by a semi-analytic galaxy formation code, SatGen. In particular, we test the effect of host galaxy halo mass, disk mass, ratio of disk scale height to length, and stellar feedback model on disrupting satellite populations. We find that the counts of tidal debris are consistent across all host galaxy models, within a given host mass range, and that all models can have stream-like debris on low-energy orbits, consistent with those observed around the Milky Way. However, we find a preference for stream-like debris on lower-energy orbits in models with a thicker (lower-density) host disk or on higher-energy orbits in models with a more-massive host disk. Importantly, we observe significant halo-to-halo variance across all models. These results highlight the importance of simulating and observing large samples of Milky Way-mass galaxies and accounting for variations in host properties when using disrupting satellites in studies of near-field cosmology.

A Field Guide to Event-Shape Observables Using Optimal Transport
Cari Cesarotti, Matt LeBlanc
[ arXiv:2409.13150 ]
Theoretical Physics

Abstract We lay out the phenomenological behavior of event-shape observables evaluated by solving optimal transport problems between collider events and reference geometries -- which we name 'manifold distances' -- to provide guidance regarding their use in future studies. This discussion considers several choices related to the metric used to quantify these distances. We explore the differences between the various options, using a combination of analytical studies and simulated minimum-bias and multi-jet events. Making judicious choices when defining the metric and reference geometry can improve sensitivity to interesting signal features and reduce sensitivity to non-perturbative effects in QCD. The goal of this article is to provide a 'field guide' that can inform how choices made when defining a manifold distance can be tailored for the analysis at-hand.

Conformal Fields from Neural Networks
James Halverson, Joydeep Naskar, Jiahua Tian
[ arXiv:2409.12222 ]
Theoretical Physics

Abstract We use the embedding formalism to construct conformal fields in D dimensions, by restricting Lorentz-invariant ensembles of homogeneous neural networks in (D+2) dimensions to the projective null cone. Conformal correlators may be computed using the parameter space description of the neural network. Exact four-point correlators are computed in a number of examples, and we perform a 4D conformal block decomposition that elucidates the spectrum. In some examples the analysis is facilitated by recent approaches to Feynman integrals. Generalized free CFTs are constructed using the infinite-width Gaussian process limit of the neural network, enabling a realization of the free boson. The extension to deep networks constructs conformal fields at each subsequent layer, with recursion relations relating their conformal dimensions and four-point functions. Numerical approaches are discussed.

Full-shape analysis with simulation-based priors: cosmological parameters and the structure growth anomaly
Mikhail M. Ivanov, Andrej Obuljen, Carolina Cuesta-Lazaro, Michael W. Toomey
Physical Review D, 2025, Vol. 111, Iss. 6 [ arXiv:2409.10609 ] published
Astrophysics

Abstract We explore full-shape analysis with simulation-based priors, which is the simplest approach to galaxy clustering data analysis that combines effective field theory (EFT) on large scales and numerical simulations on small scales. The core ingredient of our approach is the prior density of EFT parameters which we extract from a suite of 10500 galaxy simulations based on the halo occupation distribution (HOD) model. We measure the EFT parameters with the field-level forward model, which enables us to cancel cosmic variance. On the theory side, we develop a new efficient approach to calculate field-level transfer functions using time-sliced perturbation theory and the logarithmic fast Fourier transform. We study cosmology dependence of EFT parameters of dark matter halos and HOD galaxies and find that it can be ignored for the purpose of prior generation. We use neural density estimation to model the measured distribution of EFT parameters. Our distribution model is then used as a prior in a reanalysis of the BOSS full-shape galaxy power spectrum data. Assuming the ΛCDM model, we find significant (≈30% and ≈60%) improvements for the matter density fraction and the mass fluctuation amplitude, which are constrained to Ωm=0.315±0.010 and σ8=0.671±0.027. The value of the Hubble constant does not change, H0=68.7±1.1 km/s/Mpc. This reaffirms earlier reports of the structure growth tension from the BOSS data. Finally, we use the measured EFT parameters to constrain galaxy formation physics.

Auxiliary Field Deformations of (Semi-)Symmetric Space Sigma Models
Daniele Bielli, Christian Ferko, Liam Smith, Gabriele Tartaglino-Mazzucchelli
Journal of High Energy Physics, Volume 2025, article number 96 [ arXiv:2409.05704 ] published
Theoretical Physics

Abstract We generalize the auxiliary field deformations of the principal chiral model (PCM) introduced in arXiv:2405.05899 and arXiv:2407.16338 to sigma models whose target manifolds are symmetric or semi-symmetric spaces, including a Wess-Zumino term in the latter case. This gives rise to a new infinite family of classically integrable Z2 and Z4 coset models of the form which are of interest in applications of integrability to worldsheet string theory and holography. We demonstrate that every theory in this infinite class admits a zero-curvature representation for its equations of motion by exhibiting a Lax connection.

How DREAMS are made: Emulating Satellite Galaxy and Subhalo Populations with Diffusion Models and Point Clouds
Tri Nguyen, Francisco Villaescusa-Navarro, Siddharth Mishra-Sharma, Carolina Cuesta-Lazaro, Paul Torrey, Arya Farahi, Alex M. Garcia, Jonah C. Rose, Stephanie O’Neil, Mark Vogelsberger, Xuejian Shen, Cian Roche, Daniel Anglés-Alcázar, Nitya Kallivayalil, Julian B. Muñoz, Francis-Yan Cyr-Racine, Sandip Roy, Lina Necib, Kassidy E. Kollmann
[ arXiv:2409.02980 | code ]
Astrophysics

Abstract The connection between galaxies and their host dark matter (DM) halos is critical to our understanding of cosmology, galaxy formation, and DM physics. To maximize the return of upcoming cosmological surveys, we need an accurate way to model this complex relationship. Many techniques have been developed to model this connection, from Halo Occupation Distribution (HOD) to empirical and semi-analytic models to hydrodynamic. Hydrodynamic simulations can incorporate more detailed astrophysical processes but are computationally expensive; HODs, on the other hand, are computationally cheap but have limited accuracy. In this work, we present NeHOD, a generative framework based on variational diffusion model and Transformer, for painting galaxies/subhalos on top of DM with an accuracy of hydrodynamic simulations but at a computational cost similar to HOD. By modeling galaxies/subhalos as point clouds, instead of binning or voxelization, we can resolve small spatial scales down to the resolution of the simulations. For each halo, NeHOD predicts the positions, velocities, masses, and concentrations of its central and satellite galaxies. We train NeHOD on the TNG-Warm DM suite of the DREAMS project, which consists of 1024 high-resolution zoom-in hydrodynamic simulations of Milky Way-mass halos with varying warm DM mass and astrophysical parameters. We show that our model captures the complex relationships between subhalo properties as a function of the simulation parameters, including the mass functions, stellar-halo mass relations, concentration-mass relations, and spatial clustering. Our method can be used for a large variety of downstream applications, from galaxy clustering to strong lensing studies.

Double ‘acct’: a distinct double-peaked supernova matching pulsational pair-instability models
C. R. Angus, S. E. Woosley, R. J. Foley, M. Nicholl, V. A. Villar, K. Taggart, M. Pursiainen, P. Ramsden, S. Srivastav, H. F. Stevance, T. Moore, K. Auchettl, W. B. Hoogendam, N. Khetan, S. K. Yadavalli, G. Dimitriadis, A. Gagliano, M. R. Siebert, A. Aamer, T. de Boer, K. C. Chambers, A. Clocchiatti, D. A. Coulter, M. R. Drout, D. Farias, M. D. Fulton, C. Gall, H. Gao, L. Izzo, D. O. Jones, C.-C. Lin, E. A. Magnier, G. Narayan, E. Ramirez-Ruiz, C. L. Ransome, A. Rest, S. J. Smartt, K. W. Smith
[ arXiv:2409.02174 ]
Experimental Physics

Abstract We present multi-wavelength data of SN2020acct, a double-peaked stripped-envelope supernova (SN) in NGC2981 at ~150 Mpc. The two peaks are temporally distinct, with maxima separated by 58 rest-frame days, and a factor of 20 reduction in flux between. The first is luminous (Mr = -18.00 ± 0.02 mag), blue (g - r = 0.27 ± 0.03 mag), and displays spectroscopic signatures of interaction with hydrogen-free circumstellar material. The second peak is fainter (Mr = -17.29 ± 0.03 mag), and spectroscopically similar to an evolved stripped-envelope SNe, with strong blended forbidden [Ca II] and [O II] features. No other known double-peak SN exhibits a light curve similar to that of SN 2020acct. We find the likelihood of two individual SNe occurring in the same star-forming region within that time to be highly improbable, while an implausibly fine-tuned configuration would be required to produce two SNe from a single binary system. We find that the peculiar properties of SN2020acct match models of pulsational pair instability (PPI), in which the initial peak is produced by collisions of shells of ejected material, shortly followed by a terminal explosion. Pulsations from a star with a 72 M⊙ helium core provide an excellent match to the double-peaked light curve. The local galactic environment has a metallicity of 0.4 Z⊙, a level where massive single stars are not expected retain enough mass to encounter the PPI. However, late binary mergers or a low-metallicity pocket may allow the required core mass. We measure the rate of SN 2020acct-like events to be <3.3×10−8 Mpc−3 yr−1 at z = 0.07, or <0.1% of the total core-collapse SN rate.

SN 2021foa: The ‘Flip-Flop’ Type IIn / Ibn supernova
D. Farias, C. Gall, G. Narayan, S. Rest, V. A. Villar, C. R. Angus, K. Auchettl, K. W. Davis, R. Foley, A. Gagliano, J. Hjorth, L. Izzo, C. D. Kilpatrick, H .M. L. Perkins, E. Ramirez-Ruiz, C. L. Ransome, Sarangi. A., R. Yarza, D. A. Coulter, D. O. Jones, N. Khetan, A. Rest, M. R. Siebert, J. J. Swift, K. Taggart, S. Tinyanont, P. Wrubel, T. J. L. de Boer, K. E. Clever, A. Dhara, H. Gao, C.-C. Lin
[ arXiv:2409.01359 ]
Experimental Physics

Abstract We present a comprehensive analysis of the photometric and spectroscopic evolution of SN~2021foa, unique among the class of transitional supernovae for repeatedly changing its spectroscopic appearance from hydrogen-to-helium-to-hydrogen-dominated (IIn-to-Ibn-to-IIn) within 50 days past peak brightness. The spectra exhibit multiple narrow (≈ 300--600~km~s−1) absorption lines of hydrogen, helium, calcium and iron together with broad helium emission lines with a full-width-at-half-maximum (FWHM) of ∼6000~km~s−1. For a steady, wind-mass loss regime, light curve modeling results in an ejecta mass of ∼8 M⊙ and CSM mass below 1 M⊙, and an ejecta velocity consistent with the FWHM of the broad helium lines. We obtain a mass-loss rate of ≈2 M⊙yr−1. This mass-loss rate is three orders of magnitude larger than derived for normal Type II SNe. We estimate that the bulk of the CSM of SN~2021foa must have been expelled within half a year, about 15 years ago. Our analysis suggests that SN~2021foa had a helium rich ejecta which swept up a dense shell of hydrogen rich CSM shortly after explosion. At about 60 days past peak brightness, the photosphere recedes through the dense ejecta-CSM region, occulting much of the red-shifted emission of the hydrogen and helium lines, which results in observed blue-shift (∼−3000~km~s−1). Strong mass loss activity prior to explosion, such as those seen in SN~2009ip-like objects and SN~2021foa as precursor emission, are the likely origin of a complex, multiple-shell CSM close to the progenitor star.

Maven: A Multimodal Foundation Model for Supernova Science
Gemma Zhang, Thomas Helfer, Alexander T. Gagliano, Siddharth Mishra-Sharma, V. Ashley Villar
Machine Learning Science and Tehnology, Volume , Number 4, 2024 [ arXiv:2408.16829 | code ] published
Astrophysics

Abstract A common setting in astronomy is the availability of a small number of high-quality observations, and larger amounts of either lower-quality observations or synthetic data from simplified models. Time-domain astrophysics is a canonical example of this imbalance, with the number of supernovae observed photometrically outpacing the number observed spectroscopically by multiple orders of magnitude. At the same time, no data-driven models exist to understand these photometric and spectroscopic observables in a common context. Contrastive learning objectives, which have grown in popularity for aligning distinct data modalities in a shared embedding space, provide a potential solution to extract information from these modalities. We present Maven, the first foundation model for supernova science. To construct Maven, we first pre-train our model to align photometry and spectroscopy from 0.5M synthetic supernovae using a constrastive objective. We then fine-tune the model on 4,702 observed supernovae from the Zwicky Transient Facility. Maven reaches state-of-the-art performance on both classification and redshift estimation, despite the embeddings not being explicitly optimized for these tasks. Through ablation studies, we show that pre-training with synthetic data improves overall performance. In the upcoming era of the Vera C. Rubin Observatory, Maven serves as a Rosetta Stone for leveraging large, unlabeled and multimodal time-domain datasets.

Remove Symmetries to Control Model Expressivity
Liu Ziyin, Yizhou Xu, Isaac Chuang
[ arXiv:2408.15495 ]
Foundational AI

Abstract When symmetry is present in the loss function, the model is likely to be trapped in a low-capacity state that is sometimes known as a "collapse." Being trapped in these low-capacity states can be a major obstacle to training across many scenarios where deep learning technology is applied. We first prove two concrete mechanisms through which symmetries lead to reduced capacities and ignored features during training. We then propose a simple and theoretically justified algorithm, syre, to remove almost all symmetry-induced low-capacity states in neural networks. The proposed method is shown to improve the training of neural networks in scenarios when this type of entrapment is especially a concern. A remarkable merit of the proposed method is that it is model-agnostic and does not require any knowledge of the symmetry.

Low-Budget Simulation-Based Inference with Bayesian Neural Networks
Arnaud Delaunoy, Maxence de la Brassinne Bonardeaux, Siddharth Mishra-Sharma, Gilles Louppe
[ arXiv:2408.15136 | code ]
Foundational AI

Abstract Simulation-based inference methods have been shown to be inaccurate in the data-poor regime, when training simulations are limited or expensive. Under these circumstances, the inference network is particularly prone to overfitting, and using it without accounting for the computational uncertainty arising from the lack of identifiability of the network weights can lead to unreliable results. To address this issue, we propose using Bayesian neural networks in low-budget simulation-based inference, thereby explicitly accounting for the computational uncertainty of the posterior approximation. We design a family of Bayesian neural network priors that are tailored for inference and show that they lead to well-calibrated posteriors on tested benchmarks, even when as few as O(10) simulations are available. This opens up the possibility of performing reliable simulation-based inference using very expensive simulators, as we demonstrate on a problem from the field of cosmology where single simulations are computationally expensive. We show that Bayesian neural networks produce informative and well-calibrated posterior estimates with only a few hundred simulations.

Cosmological Parameter Estimation with a Joint-Likelihood Analysis of the Cosmic Microwave Background and Big Bang Nucleosynthesis
Cara Giovanetti, Mariangela Lisanti, Hongwan Liu, Siddharth Mishra-Sharma, Joshua T. Ruderman
[ arXiv:2408.14531 ]
Astrophysics

Abstract We present the first joint-likelihood analysis of Big Bang Nucleosynthesis (BBN) and Cosmic Microwave Background (CMB) data. Bayesian inference is performed on the baryon abundance and the effective number of neutrino species, Neff, using a CMB Boltzmann solver in combination with LINX, a new flexible and efficient BBN code. We marginalize over Planck nuisance parameters and nuclear rates to find Neff=3.08+0.13−0.13,2.94+0.16−0.16, or 2.98+0.14−0.13, for three separate reaction networks. This framework enables robust testing of the Lambda Cold Dark Matter paradigm and its variants with CMB and BBN data.

LINX: A Fast, Differentiable, and Extensible Big Bang Nucleosynthesis Package
Cara Giovanetti, Mariangela Lisanti, Hongwan Liu, Siddharth Mishra-Sharma, Joshua T. Ruderman
[ arXiv:2408.14538 | code ]
Astrophysics

Abstract We introduce LINX (Light Isotope Nucleosynthesis with JAX), a new differentiable public Big Bang Nucleosynthesis (BBN) code designed for fast parameter estimation. By leveraging JAX, LINX achieves both speed and differentiability, enabling the use of Bayesian inference, including gradient-based methods. We discuss the formalism used in LINX for rapid primordial elemental abundance predictions and give examples of how LINX can be used. When combined with differentiable Cosmic Microwave Background (CMB) power spectrum emulators, LINX can be used for joint CMB and BBN analyses without requiring extensive computational resources, including on personal hardware.

Finding the Fuse: Prospects for the Detection and Characterization of Hydrogen-Rich Core-Collapse Supernova Precursor Emission with the LSST
A. Gagliano, E. Berger, V. A. Villar, D. Hiramatsu, R. Kessler, T. Matsumoto, A. Gilkis, E. Laplace
The Astrophysical Journal, Volume 978, Number 1, 2024 [ arXiv:2408.13314 ] published
Experimental Physics

Abstract Enhanced emission in the months to years preceding explosion has been detected for several core-collapse supernovae (SNe). Though the physical mechanisms driving the emission remain hotly debated, the light curves of detected events show long-lived (≥50 days), plateau-like behavior, suggesting hydrogen recombination may significantly contribute to the total energy budget. The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) will provide a decade-long photometric baseline to search for this emission, both in binned pre-explosion observations after an SN is detected and in single-visit observations prior to the SN explosion. In anticipation of these searches, we simulate a range of eruptive precursor models to core-collapse SNe and forecast the discovery rates of these phenomena in LSST data. We find a detection rate of ~40-130 yr−1 for SN IIP/IIL precursors and ~110 yr−1 for SN IIn precursors in single-epoch photometry. Considering the first three years of observations with the effects of rolling and observing triplets included, this number grows to a total of 150-400 in binned photometry, with the highest number recovered when binning in 100-day bins for 2020tlf-like precursors and in 20-day bins for other recombination-driven models from the literature. We quantify the impact of using templates contaminated by residual light (from either long-lived or separate precursor emission) on these detection rates, and explore strategies for estimating baseline flux to mitigate these issues. Spectroscopic follow-up of the eruptions preceding core-collapse SNe and detected with LSST will offer important clues to the underlying drivers of terminal-stage mass loss in massive stars.

Multiple testing for signal-agnostic searches of new physics with machine learning
Gaia Grosso, Marco Letizia
The European Physical Journal, Volume 85, article number 4, 2025 [ arXiv:2408.12296 | code ] published
Experimental Physics

Abstract In this work, we address the question of how to enhance signal-agnostic searches by leveraging multiple testing strategies. Specifically, we consider hypothesis tests relying on machine learning, where model selection can introduce a bias towards specific families of new physics signals. We show that it is beneficial to combine different tests, characterised by distinct choices of hyperparameters, and that performances comparable to the best available test are generally achieved while providing a more uniform response to various types of anomalies. Focusing on the New Physics Learning Machine, a methodology to perform a signal-agnostic likelihood-ratio test, we explore a number of approaches to multiple testing, such as combining p-values and aggregating test statistics.

Quantum inception score
Akira Sone, Akira Tanji, and Naoki Yamamoto
Physical Review Research, 2024, Volume 6, Issue 3 [ arXiv:2311.12163 ] published
Foundational AI

Abstract Motivated by the great success of classical generative models in machine learning, enthusiastic exploration of their quantum version has recently started. To depart on this journey, it is important to develop a relevant metric to evaluate the quality of quantum generative models; in the classical case, one such example is the (classical) inception score (cIS). In this paper, as a natural extension of cIS, we propose the quantum inception score (qIS) for quantum generators. Importantly, qIS relates the quality to the Holevo information of the quantum channel that classifies a given dataset. In this context, we show several properties of qIS. First, qIS is greater than or equal to the corresponding cIS, which is defined through projection measurements on the system output. Second, the difference between qIS and cIS arises from the presence of quantum coherence, as characterized by the resource theory of asymmetry. Third, when a set of entangled generators is prepared, there exists a classifying process leading to the further enhancement of qIS. Fourth, we harness the quantum fluctuation theorem to characterize the physical limitation of qIS. Finally, we apply qIS to assess the quality of the one-dimensional spin chain model as a quantum generative model, with the quantum convolutional neural network as a quantum classifier, for the phase classification problem in the quantum many-body physics.

KAN 2.0: Kolmogorov-Arnold Networks Meet Science
Ziming Liu, Pingchuan Ma, Yixuan Wang, Wojciech Matusik, Max Tegmark
[ arXiv:2408.10205 | code ]
Foundational AI

Abstract A major challenge of AI + Science lies in their inherent incompatibility: today's AI is primarily based on connectionism, while science depends on symbolism. To bridge the two worlds, we propose a framework to seamlessly synergize Kolmogorov-Arnold Networks (KANs) and science. The framework highlights KANs' usage for three aspects of scientific discovery: identifying relevant features, revealing modular structures, and discovering symbolic formulas. The synergy is bidirectional: science to KAN (incorporating scientific knowledge into KANs), and KAN to science (extracting scientific insights from KANs). We highlight major new functionalities in the pykan package: (1) MultKAN: KANs with multiplication nodes. (2) kanpiler: a KAN compiler that compiles symbolic formulas into KANs. (3) tree converter: convert KANs (or any neural networks) to tree graphs. Based on these tools, we demonstrate KANs' capability to discover various types of physical laws, including conserved quantities, Lagrangians, symmetries, and constitutive laws.

Enhancing events in neutrino telescopes through deep-learning-driven superresolution
Felix J. Yu, Nicholas Kamp, Carlos A. Argüelles
Physical Review D, 2025, Volume 111, Issue 4 [ arXiv:2408.08474 | code ] published
Experimental Physics

Abstract Recent discoveries by neutrino telescopes, such as the IceCube Neutrino Observatory, relied extensively on machine learning (ML) tools to infer physical quantities from the raw photon hits detected. Neutrino telescope reconstruction algorithms are limited by the sparse sampling of photons by the optical modules due to the relatively large spacing (10−100m) between them. In this letter, we propose a novel technique that learns photon transport through the detector medium through the use of deep learning-driven super-resolution of data events. These ``improved'' events can then be reconstructed using traditional or ML techniques, resulting in improved resolution. Our strategy arranges additional ``virtual'' optical modules within an existing detector geometry and trains a convolutional neural network to predict the hits on these virtual optical modules. We show that this technique improves the angular reconstruction of muons in a generic ice-based neutrino telescope. Our results readily extend to water-based neutrino telescopes and other event morphologies.

Improving Radial Velocities by Marginalizing over Stars and Sky: Achieving 30 m/s RV Precision for APOGEE in the Plate Era
Andrew K. Saydjari, Douglas P. Finkbeiner, Adam J. Wheeler, Jon A. Holtzman, John C. Wilson, Andrew R. Casey, Sophia Sánchez-Maes, Joel R. Brownstein, David W. Hogg, Michael R. Blanton
[ arXiv:2408.07126 | code ]
Astrophysics

Abstract The radial velocity catalog from the Apache Point Observatory Galactic Evolution Experiment (APOGEE) is unique in its simultaneously large volume and high precision as a result of its decade-long survey duration, multiplexing (600 fibers), and spectral resolution of R∼22,500. However, previous data reductions of APOGEE have not fully realized the potential radial velocity (RV) precision of the instrument. Here we present an RV catalog based on a new reduction of all 2.6 million visits of APOGEE DR17 and validate it against improved estimates for the theoretical RV performance. The core ideas of the new reduction are the simultaneous modeling of all components in the spectra, rather than a separate subtraction of point estimates for the sky, and a marginalization over stellar types, rather than a grid search for an optimum. We show that this catalog, when restricted to RVs measured with the same fiber, achieves noise-limited precision down to 30 m/s and delivers well-calibrated uncertainties. We also introduce a general method for calibrating fiber-to-fiber constant RV offsets and demonstrate its importance for high RV precision work in multi-fiber spectrographs. After calibration, we achieve 47 m/s RV precision on the combined catalog with RVs measured with different fibers. This degradation in precision relative to measurements with only a single fiber suggests that refining line spread function models should be a focus in SDSS-V to improve the fiber-unified RV catalog.

Learning the Simplicity of Scattering Amplitudes
Clifford Cheung, Aurélien Dersy, Matthew D. Schwartz
SciPost Physics, 2025, Vol.18, Issue 2 [ arXiv:2408.04720 | code ] published
Theoretical Physics

Abstract The simplification and reorganization of complex expressions lies at the core of scientific progress, particularly in theoretical high-energy physics. This work explores the application of machine learning to a particular facet of this challenge: the task of simplifying scattering amplitudes expressed in terms of spinor-helicity variables. We demonstrate that an encoder-decoder transformer architecture achieves impressive simplification capabilities for expressions composed of handfuls of terms. Lengthier expressions are implemented in an additional embedding network, trained using contrastive learning, which isolates subexpressions that are more likely to simplify. The resulting framework is capable of reducing expressions with hundreds of terms - a regular occurrence in quantum field theory calculations - to vastly simpler equivalent expressions. Starting from lengthy input expressions, our networks can generate the Parke-Taylor formula for five-point gluon scattering, as well as new compact expressions for five-point amplitudes involving scalars and gravitons. An interactive demonstration can be found at this https URL.

Hyperspectral Multilevel GCN and CNN Feature Fusion for Change Detection
Chhaya Katiyar, Vidya Manian
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024 [ ] published
Foundational AI

Abstract Hyperspectral Image Change Detection (HSI-CD) focuses on identifying differences in multi-temporal HSIs. Graph Convolutional Networks (GCN) have demonstrated greater promise than convolutional neural networks (CNNs) in remote sensing, particularly for processing HSIs. This is due to GCN's ability to handle non-Euclidean graph-structured information, as opposed to the fixed kernel operations of CNN based on Euclidean structures. Specifically, GCN operates predominantly on superpixel-based nodes. This paper proposes a method, named Hyperspectral Multilevel GCN and CNN Feature Fusion for Change Detection (HMGCF-CD), that integrates superpixel-level GCN with pixel-level CNN for feature extraction and efficient change detection in HSIs. The proposed method utilizes the strengths of both CNN and GCN; the CNN branch focuses on feature learning in small-scale, regular regions, while the GCN branch handles large-scale, irregular regions. This approach generates complementary spectral-spatial features at both pixel and superpixel levels. To bridge the structural incompatibility between the Euclidean-data-oriented CNN and the non-Euclideandata-oriented GCN, HMGCF introduces a graph encoder and decoder. These elements help in propagating features between image pixels and graph nodes, allowing CNN and GCN to function within an integrated end-to-end framework. HMGCF integrates graph encoding into the network, edge weights, and node representations from training data. This innovation improves node feature learning and adapts the graph more effectively to HSI content. Fusion of features from different branches within the HMGCF enables precise changes in HSI. Experiments carried out on four datasets demonstrate that the proposed HMGCF method outperforms existing methods, marking a significant advancement in HSI-CD techniques.

AT2023vto: An Exceptionally Luminous Helium Tidal Disruption Event from a Massive Star
Harsh Kumar, Edo Berger, Daichi Hiramatsu, Sebastian Gomez, Peter K. Blanchard, Yvette Cendes, K. Azalee Bostroem, Joseph Farah, Estefania Padilla Gonzalez, Andrew Howell, Curtis McCully, Megan Newsome, Giacomo Terreran
[ arXiv:2408.01482 ]
Astrophysics

Abstract We present optical/UV observations and the spectroscopic classification of the transient AT2023vto as a tidal disruption event (TDE) at z = 0.4846. The spectrum is dominated by a broad He II λ4686 emission line, with a width of ~ 3.76×104 km/s and a blueshift of ~ 1.05×104 km/s, classifying it as a member of the TDE-He class. The light curve exhibits a long rise and decline timescale, with a large peak absolute magnitude of Mg ~ -23.6, making it the most luminous of the classical optical TDEs (H, H+He, He) discovered to date by about 2 mag (and ~ 4 mag compared to the mean of the population). The light curve exhibits a persistent blue color of g - r ~ -0.4 mag throughout its evolution, similar to other TDEs, but distinct from supernovae. We identify the host galaxy of AT2023vto in archival Pan-STARRS images and find that the transient is located at the galaxy center, and that its inferred central black hole mass is ~ 107 M⊙. Modeling the light curves of AT2023vto, we find that it resulted from the disruption of a ~ 9 M⊙ star by a ~107 M⊙ supermassive black hole. The star mass is about 5 times larger than the highest star masses previously inferred in TDEs, and the black hole mass is at the high end of the distribution. AT2023vto is comparable in luminosity and timescale to some putative TDEs (with a blue featureless continuum), as well as to the mean of the recently identified population of ambiguous nuclear transients (ANTs), although the latter are spectroscopically distinct and tend to have longer timescales. ANTs have been speculated to arise from tidal disruptions of massive stars, perhaps in active galactic nuclei, and AT2023vto may represent a similar case but in a dormant black hole, thereby bridging the TDE and ANT populations. We anticipate that Rubin Observatory / LSST will uncover similar luminous TDEs to z ~ 3.

Attractors, Geodesics, and the Geometry of Moduli Spaces
Fabian Ruehle, Benjamin Sung
[ arXiv:2408.00830 ]
Theoretical Physics

Abstract We connect recent conjectures and observations pertaining to geodesics, attractor flows, Laplacian eigenvalues and the geometry of moduli spaces by using that attractor flows are geodesics. For toroidal compactifications, attractor points are related to (degenerate) masses of the Laplacian on the target space, and also to the Laplacian on the moduli space. We also explore compactifications of M-Theory to 5D on a Calabi-Yau threefold and argue that geodesics are unique in a special set of classes, providing further evidence for a recent conjecture by Raman and Vafa. Finally, we describe the role of the marked moduli space in 4d =2 compactifications. We study split attractor flows in an explicit example of the one-parameter family of quintics and discuss setups where flops to isomorphic Calabi-Yau manifolds exist.

On the Generality and Persistence of Cosmological Stasis
James Halverson, Sneh Pandya
Physical Review D, Vol. 110, Iss. 7, 2024 [ arXiv:2408.00835 | code ] published
Astrophysics

Abstract Hierarchical decays of N matter species to radiation may balance against Hubble expansion to yield stasis, a new phase of cosmological evolution with constant matter and radiation abundances. We analyze stasis with various machine learning techniques on the full 2N-dimensional space of decay rates and abundances, which serve as inputs to the system of Boltzmann equations that governs the dynamics. We construct a differentiable Boltzmann solver to maximize the number of stasis e-folds . High-stasis configurations obtained by gradient ascent motivate log-uniform distributions on rates and abundances to accompany power-law distributions of previous works. We demonstrate that random configurations drawn from these families of distributions regularly exhibit many e-folds of stasis. We additionally use them as priors in a Bayesian analysis conditioned on stasis, using stochastic variational inference with normalizing flows to model the posterior. All three numerical analyses demonstrate the generality of stasis and point to a new model in which the rates and abundances are exponential in the species index. We show that the exponential model solves the exact stasis equations, is an attractor, and satisfies ∝N, exhibiting inflation-level e-folding with a relatively low number of species. This is contrasted with the ∝log(N) scaling of power-law models. Finally, we discuss implications for the emergent string conjecture and string axiverse.

TASI Lectures on Physics for Machine Learning
Jim Halverson
[ arXiv:2408.00082 ]
Theoretical Physics

Abstract These notes are based on lectures I gave at TASI 2024 on Physics for Machine Learning. The focus is on neural network theory, organized according to network expressivity, statistics, and dynamics. I present classic results such as the universal approximation theorem and neural network / Gaussian process correspondence, and also more recent results such as the neural tangent kernel, feature learning with the maximal update parameterization, and Kolmogorov-Arnold networks. The exposition on neural network theory emphasizes a field theoretic perspective familiar to theoretical physicists. I elaborate on connections between the two, including a neural network approach to field theory.

Relaxed Equivariant Graph Neural Networks
Elyssa Hofgard, Rui Wang, Robin Walters, Tess Smidt
[ arXiv:2407.20471 | code ]
Foundational AI

Abstract 3D Euclidean symmetry equivariant neural networks have demonstrated notable success in modeling complex physical systems. We introduce a framework for relaxed E(3) graph equivariant neural networks that can learn and represent symmetry breaking within continuous groups. Building on the existing e3nn framework, we propose the use of relaxed weights to allow for controlled symmetry breaking. We show empirically that these relaxed weights learn the correct amount of symmetry breaking.

A Generative Modeling Approach to Reconstructing 21-cm Tomographic Data
Nashwan Sabti, Ram Reddy, Julian B. Muñoz, Siddharth Mishra-Sharma, Taewook Youn
Machine Learning: Science and Technology, 2025, Volume 6, Number 1 [ arXiv:2407.21097 | code ] published
Astrophysics

Abstract Analyses of the cosmic 21-cm signal are hampered by astrophysical foregrounds that are far stronger than the signal itself. These foregrounds, typically confined to a wedge-shaped region in Fourier space, often necessitate the removal of a vast majority of modes, thereby degrading the quality of the data anisotropically. To address this challenge, we introduce a novel deep generative model based on stochastic interpolants to reconstruct the 21-cm data lost to wedge filtering. Our method leverages the non-Gaussian nature of the 21-cm signal to effectively map wedge-filtered 3D lightcones to samples from the conditional distribution of wedge-recovered lightcones. We demonstrate how our method is able to restore spatial information effectively, considering both varying cosmological initial conditions and astrophysics. Furthermore, we discuss a number of future avenues where this approach could be applied in analyses of the 21-cm signal, potentially offering new opportunities to improve our understanding of the Universe during the epochs of cosmic dawn and reionization.

Automatic Environment Shaping is the Next Frontier in RL
Younghyo Park, Gabriel B. Margolis, Pulkit Agrawal
[ arXiv:2407.16186 ]
Foundational AI

Abstract Many roboticists dream of presenting a robot with a task in the evening and returning the next morning to find the robot capable of solving the task. What is preventing us from achieving this? Sim-to-real reinforcement learning (RL) has achieved impressive performance on challenging robotics tasks, but requires substantial human effort to set up the task in a way that is amenable to RL. It's our position that algorithmic improvements in policy optimization and other ideas should be guided towards resolving the primary bottleneck of shaping the training environment, i.e., designing observations, actions, rewards and simulation dynamics. Most practitioners don't tune the RL algorithm, but other environment parameters to obtain a desirable controller. We posit that scaling RL to diverse robotic tasks will only be achieved if the community focuses on automating environment shaping procedures.

Moment Unfolding
Krish Desai, Benjamin Nachman, Jesse Thaler
Physical Review D, Vol. 110, Iss. 11, 2024 [ arXiv:2407.11284 | code ] published
Experimental Physics

Abstract Deconvolving ('unfolding') detector distortions is a critical step in the comparison of cross section measurements with theoretical predictions in particle and nuclear physics. However, most existing approaches require histogram binning while many theoretical predictions are at the level of statistical moments. We develop a new approach to directly unfold distribution moments as a function of another observable without having to first discretize the data. Our Moment Unfolding technique uses machine learning and is inspired by Generative Adversarial Networks (GANs). We demonstrate the performance of this approach using jet substructure measurements in collider physics. With this illustrative example, we find that our Moment Unfolding protocol is more precise than bin-based approaches and is as or more precise than completely unbinned methods.

The Type I Superluminous Supernova Catalog I: Light Curve Properties, Models, and Catalog Description
Sebastian Gomez, Matt Nicholl, Edo Berger, Peter K. Blanchard, V. Ashley Villar, Sofia Rest, Griffin Hosseinzadeh, Aysha Aamer, Yukta Ajay, Wasundara Athukoralalage, David C. Coulter, Tarraneh Eftekhari, Achille Fiore, Noah Franz, Ori Fox, Alexander Gagliano, Daichi Hiramatsu, D. Andrew Howell, Brian Hsu, Mitchell Karmen, Matthew R. Siebert, Réka Könyves-Tóth, Harsh Kumar, Curtis McCully, Craig Pellegrino, Justin Pierel, Armin Rest, Qinan Wang
Monthly Notices of the Royal Astronomical Society, 2024 [ arXiv:2407.07946 | code ] published
Astrophysics

Abstract We present the most comprehensive catalog to date of Type I Superluminous Supernovae (SLSNe), a class of stripped envelope supernovae (SNe) characterized by exceptionally high luminosities. We have compiled a sample of 262 SLSNe reported through 2022 December 31. We verified the spectroscopic classification of each SLSN and collated an exhaustive data set of UV, optical and IR photometry from both publicly available data and our own FLEET observational follow-up program, totaling over 30,000 photometric detections. Using these data we derive observational parameters such as the peak absolute magnitudes, rise and decline timescales, as well as bolometric luminosities, temperature and photospheric radius evolution for all SLSNe. Additionally, we model all light curves using a hybrid model that includes contributions from both a magnetar central engine and the radioactive decay of 56Ni. We explore correlations among various physical and observational parameters, and recover the previously found relation between ejecta mass and magnetar spin, as well as the overall progenitor pre-explosion mass distribution with a peak at ≈6.5 M⊙. We find no significant redshift dependence for any parameter, and no evidence for distinct sub-types of SLSNe. We find that <3% of SLSNe are best fit with a significant contribution from radioactive decay ≳50%, representing a set of relatively dim and slowly declining SNe. We provide several analytical tools designed to simulate typical SLSN light curves across a broad range of wavelengths and phases, enabling accurate K-corrections, bolometric scaling calculations, and inclusion of SLSNe in survey simulations or future comparison works. The complete catalog, including all of the photometry, models, and derived parameters, is made available as an open-source resource on GitHub.

Find the haystacks, then look for needles: The rate of strongly lensed transients in galaxy-galaxy strong gravitational lenses
Ana Sainz de Murieta, Thomas E. Collett, Mark R. Magee, Justin D. R. Pierel, Wolfgang J. R. Enzi, Martine Lokken, Alex Gagliano, Dan Ryczanowski
[ arXiv:2407.04080 ]
Astrophysics

Abstract The time delay between appearances of multiple images of a gravitationally lensed supernova (glSN) is sensitive to the Hubble constant, H0. As well as time delays, a lensed host galaxy is needed to enable precise inference of H0. In this work we investigate the connection between discoverable lensed transients and their host galaxies. We find that LSST will discover 88 glSNe per year, of which 54% will also have a strongly lensed host. The rates can change by approximately 30 percent uncertainty depending primarily on the choice of unlensed SN population and uncertainties in the redshift evolution of the deflector population, but the fraction of glSNe with a lensed host is consistently around a half. LSST will discover 20 glSNe per year in systems that could plausibly have been identified by Euclid as galaxy-galaxy lenses before the discovery of the glSN. Such systems have preferentially longer time delays and therefore are well suited for cosmography. We define a golden sample of glSNe Ia with time delays over 10 days, image separations greater than 0.8 arcseconds, and a multiply imaged host. For this golden sample, we find 91% occur in systems that should already be discoverable as galaxy-galaxy lenses in Euclid. For cosmology with glSNe, monitoring Euclid lenses is a plausible alternative to searching the entire LSST alert stream.

A multicategory jet image classification framework using deep neural network
Jairo Orozco Sandoval, Vidya Manian, Sudhir Malik
[ arXiv:2407.03524 ]
Foundational AI

Abstract Jet point cloud images are high dimensional data structures that needs to be transformed to a separable feature space for machine learning algorithms to distinguish them with simple decision boundaries. In this article, the authors focus on jet category separability by particle and jet feature extraction, resulting in more efficient training of a simple deep neural network, resulting in a computational efficient interpretable model for jet classification. The methodology is tested with three to five categories of jets from the JetNet benchmark jet tagging dataset, resulting in comparable performance to particle flow network. This work demonstrates that high dimensional datasets represented in separable latent spaces lead to simpler architectures for jet classification.

Anomaly-aware summary statistic from data batches
Gaia Grosso
Journal of High Energy Physics, Volume 2024, article number 93, 2024 [ arXiv:2407.01249 ] published
Experimental Physics

Abstract Signal-agnostic data exploration based on machine learning could unveil very subtle statistical deviations of collider data from the expected Standard Model of particle physics. The beneficial impact of a large training sample on machine learning solutions motivates the exploration of increasingly large and inclusive samples of acquired data with resource efficient computational methods. In this work we consider the New Physics Learning Machine (NPLM), a multivariate goodness-of-fit test built on the Neyman-Pearson maximum-likelihood-ratio construction, and we address the problem of testing large size samples under computational and storage resource constraints. We propose to perform parallel NPLM routines over batches of the data, and to combine them by locally aggregating over the data-to-reference density ratios learnt by each batch. The resulting data hypothesis defining the likelihood-ratio test is thus shared over the batches, and complies with the assumption that the expected rate of new physical processes is time invariant. We show that this method outperforms the simple sum of the independent tests run over the batches, and can recover, or even surpass, the sensitivity of the single test run over the full data. Beside the significant advantage for the offline application of NPLM to large size samples, the proposed approach offers new prospects toward the use of NPLM to construct anomaly-aware summary statistics in quasi-online data streaming scenarios.

Boosting Soft Q-Learning by Bounding
Jacob Adamczyk, Volodymyr Makarenko, Stas Tiomkin, Rahul V. Kulkarni
Reinforcement Learning Journal, Volume 5, 2024, pp. 2373–2399 [ arXiv:2406.18033 | code ] published
Foundational AI

Abstract An agent's ability to leverage past experience is critical for efficiently solving new tasks. Prior work has focused on using value function estimates to obtain zero-shot approximations for solutions to a new task. In soft Q-learning, we show how any value function estimate can also be used to derive double-sided bounds on the optimal value function. The derived bounds lead to new approaches for boosting training performance which we validate experimentally. Notably, we find that the proposed framework suggests an alternative method for updating the Q-function, leading to boosted performance.

Simulating moiré quantum matter with neural network
Di Luo, David D. Dai, Liang Fu
[ arXiv:2406.17645 ]
Theoretical Physics

Abstract Moiré materials provide an ideal platform for exploring quantum phases of matter. However, solving the many-electron problem in moiré systems is challenging due to strong correlation effects. We introduce a powerful variational representation of quantum states, many-body neural Bloch wavefunction, to solve many-electron problems in moiré materials accurately and efficiently. Applying our method to the semiconductor heterobilayer WSe2/WS2 , we obtain a generalized Wigner crystal at filling factor n = 1/3, a Mott insulator n = 1, and a correlated insulator with local magnetic moments and antiferromagnetic spin correlation at n = 2. Our neural network approach improves the simulation accuracy of strongly interacting moiré materials and paves the way for discovery of new quantum phases with variational learning principle in a unified framework.

Constructing gauge-invariant neural networks for scientific applications
Manos Theodosis, Demba E. Ba, Nima Dehmamy
ICML 2024 Workshop AI4Science [ | code ] published
Foundational AI

Abstract Our current models for fundamental forces in nature are “gauge theories”. These models are suitable for systems where interactions are local and where the local choice of coordinates does not affect physical quantities. While recent works have introduced gauge equivariant neural networks, these models focus on tangent bundles or quotient space and are not applicable to most gauge theories appearing in physics. We propose an architecture for learning general gauge invariant quantities. Our framework fills a gap in the existing literature, providing a general recipe for gauge invariance without restrictions on the spaces of the measurement vectors. We evaluate our method on a classical physical system, the XY model, that is invariant to the choice of local gauges.

Demonstration of neutron identification in neutrino interactions in the MicroBooNE liquid argon time projection chamber
MicroBooNE collaboration: P. Abratenko, O. Alterkait, D. Andrade Aldana, L. Arellano, J. Asaadi, A. Ashkenazi, S. Balasubramanian, B. Baller, A. Barnard, G. Barr, D. Barrow, J. Barrow, V. Basque, J. Bateman, O. Benevides Rodrigues, S. Berkman, A. Bhanderi, A. Bhat, M. Bhattacharya, M. Bishai, A. Blake, B. Bogart, T. Bolton, J.Y. Book, M.B. Brunetti, L. Camilleri, Y. Cao, D. Caratelli, F. Cavanna, G. Cerati, A. Chappell, Y. Chen, J.M. Conrad, M. Convery, L. Cooper-Troendle, J.I. Crespo-Anadon, R. Cross, M. Del Tutto, S.R. Dennis, P. Detje, R. Diurba, Z. Djurcic, R. Dorrill, K. Duffy, S. Dytman, B. Eberly, P. Englezos, A. Ereditato, J.J. Evans, R. Fine, B.T. Fleming, W. Foreman, D. Franco, A.P. Furmanski, F. Gao, D. Garcia-Gamez, S. Gardiner, G. Ge, S. Gollapinni, E. Gramellini, P. Green, H. Greenlee, L. Gu, W. Gu, R. Guenette, P. Guzowski, L. Hagaman, M. D. Handley, O. Hen, C. Hilgenberg, G.A. Horton-Smith, Z. Imani, B. Irwin, M.S. Ismail, C. James, X. Ji, J.H. Jo, R.A. Johnson, Y.J. Jwa, D. Kalra, N. Kamp, G. Karagiorgi, W. Ketchum, M. Kirby, T. Kobilarcik, I. Kreslo, N. Lane, J.-Y. Li, Y. Li, K. Lin, B.R. Littlejohn, H. Liu, W.C. Louis, X. Luo, C. Mariani, D. Marsden, J. Marshall, N. Martinez, D.A. Martinez Caicedo et al.
The European Physical Journal C, Volume 84, article number 1052, 2024 [ arXiv:2406.10583 ] published
Experimental Physics

Abstract A significant challenge in measurements of neutrino oscillations is reconstructing the incoming neutrino energies. While modern fully-active tracking calorimeters such as liquid argon time projection chambers in principle allow the measurement of all final state particles above some detection threshold, undetected neutrons remain a considerable source of missing energy with little to no data constraining their production rates and kinematics. We present the first demonstration of tagging neutrino-induced neutrons in liquid argon time projection chambers using secondary protons emitted from neutron-argon interactions in the MicroBooNE detector. We describe the method developed to identify neutrino-induced neutrons and demonstrate its performance using neutrons produced in muon-neutrino charged current interactions. The method is validated using a small subset of MicroBooNE's total dataset. The selection yields a sample with 60% of selected tracks corresponding to neutron-induced secondary protons.

Improving neutrino energy estimation of charged-current interaction events with recurrent neural networks in MicroBooNE
MicroBooNE collaboration: P. Abratenko, O. Alterkait, D. Andrade Aldana, L. Arellano, J. Asaadi, A. Ashkenazi, S. Balasubramanian, B. Baller, A. Barnard, G. Barr, D. Barrow, J. Barrow, V. Basque, J. Bateman, O. Benevides Rodrigues, S. Berkman, A. Bhanderi, A. Bhat, M. Bhattacharya, M. Bishai, A. Blake, B. Bogart, T. Bolton, J.Y. Book, M.B. Brunetti, L. Camilleri, Y. Cao, D. Caratelli, F. Cavanna, G. Cerati, A. Chappell, Y. Chen, J.M. Conrad, M. Convery, L. Cooper-Troendle, J.I. Crespo-Anadon, R. Cross, M. Del Tutto, S.R. Dennis, P. Detje, R. Diurba, Z. Djurcic, R. Dorrill, K. Duffy, S. Dytman, B. Eberly, P. Englezos, A. Ereditato, J.J. Evans, R. Fine, B.T. Fleming, W. Foreman, D. Franco, A.P. Furmanski, F. Gao, D. Garcia-Gamez, S. Gardiner, G. Ge, S. Gollapinni, E. Gramellini, P. Green, H. Greenlee, L. Gu, W. Gu, R. Guenette, P. Guzowski, L. Hagaman, O. Hen, C. Hilgenberg, G.A. Horton-Smith, Z. Imani, B. Irwin, M.S. Ismail, C. James, X. Ji, J.H. Jo, R.A. Johnson, Y.J. Jwa, D. Kalra, N. Kamp, G. Karagiorgi, W. Ketchum, M. Kirby, T. Kobilarcik, I. Kreslo, N. Lane, I. Lepetic, J.-Y. Li, Y. Li, K. Lin, B.R. Littlejohn, H. Liu, W.C. Louis, X. Luo, C. Mariani, D. Marsden, J. Marshall, N. Martinez, D.A. Martinez Caicedo et al.
Physical Review D, 2024, Volume 110, Issue 9 [ arXiv:2406.10123 | code ] published
Experimental Physics

Abstract We present a deep learning-based method for estimating the neutrino energy of charged-current neutrino-argon interactions. We employ a recurrent neural network (RNN) architecture for neutrino energy estimation in the MicroBooNE experiment, utilizing liquid argon time projection chamber (LArTPC) detector technology. Traditional energy estimation approaches in LArTPCs, which largely rely on reconstructing and summing visible energies, often experience sizable biases and resolution smearing because of the complex nature of neutrino interactions and the detector response. The estimation of neutrino energy can be improved after considering the kinematics information of reconstructed final-state particles. Utilizing kinematic information of reconstructed particles, the deep learning-based approach shows improved resolution and reduced bias for the muon neutrino Monte Carlo simulation sample compared to the traditional approach. In order to address the common concern about the effectiveness of this method on experimental data, the RNN-based energy estimator is further examined and validated with dedicated data-simulation consistency tests using MicroBooNE data. We also assess its potential impact on a neutrino oscillation study after accounting for all statistical and systematic uncertainties and show that it enhances physics sensitivity. This method has good potential to improve the performance of other physics analyses.

QCD constraints on isospin-dense matter and the nuclear equation of state
Ryan Abbott, William Detmold, Marc Illa, Assumpta Parreño, Robert J. Perry, Fernando Romero-López, Phiala E. Shanahan, Michael L. Wagman
Physical Review Letters, 2025, Vol.134, Iss.1 [ arXiv:2406.09273 ] published
Theoretical Physics

Abstract Understanding the behavior of dense hadronic matter is a central goal in nuclear physics as it governs the nature and dynamics of astrophysical objects such as supernovae and neutron stars. Because of the non-perturbative nature of quantum chromodynamics (QCD), little is known rigorously about hadronic matter in these extreme conditions. Here, lattice QCD calculations are used to compute thermodynamic quantities and the equation of state of QCD over a wide range of isospin chemical potentials. Agreement is seen with chiral perturbation theory predictions when the chemical potential is small. Comparison to perturbative QCD calculations at large chemical potential allows for an estimate of the gap in the superconducting phase, and this quantity is seen to agree with perturbative determinations. Since the partition function for an isospin chemical potential, μI, bounds the partition function for a baryon chemical potential μB=3μI/2, these calculations also provide rigorous non-perturbative QCD bounds on the symmetric nuclear matter equation of state over a wide range of baryon densities for the first time.

An Elliptic Kernel Unsupervised Autoencoder-Graph Convolutional Network Ensemble Model for Hyperspectral Unmixing
Estefania Alfaro-Mejia, Carlos J Delgado, Vidya Manian
[ arXiv:2406.06742 ]
Foundational AI

Abstract Spectral Unmixing is an important technique in remote sensing used to analyze hyperspectral images to identify endmembers and estimate abundance maps. Over the past few decades, performance of techniques for endmember extraction and fractional abundance map estimation have significantly improved. This article presents an ensemble model workflow called Autoencoder Graph Ensemble Model (AEGEM) designed to extract endmembers and fractional abundance maps. An elliptical kernel is applied to measure spectral distances, generating the adjacency matrix within the elliptical neighborhood. This information is used to construct an elliptical graph, with centroids as senders and remaining pixels within the geometry as receivers. The next step involves stacking abundance maps, senders, and receivers as inputs to a Graph Convolutional Network, which processes this input to refine abundance maps. Finally, an ensemble decision-making process determines the best abundance maps based on root mean square error metric. The proposed AEGEM is assessed with benchmark datasets such as Samson, Jasper, and Urban, outperforming results obtained by baseline algorithms. For the Samson dataset, AEGEM excels in three abundance maps: water, tree and soil yielding values of 0.081, 0.158, and 0.182, respectively. For the Jasper dataset, results are improved for the tree and water endmembers with values of 0.035 and 0.060 in that order, as well as for the mean average of the spectral angle distance metric 0.109. For the Urban dataset, AEGEM outperforms previous results for the abundance maps of roof and asphalt, achieving values of 0.135 and 0.240, respectively. Additionally, for the endmembers of grass and roof, AEGEM achieves values of 0.063 and 0.094.

Separating the ‘Chirp’ from the ‘Chat’: Self-supervised Visual Grounding of Sound and Language
Mark Hamilton, Andrew Zisserman, John R. Hershey, William T. Freeman
[ arXiv:2406.05629 | code ]
Foundational AI

Abstract We present DenseAV, a novel dual encoder grounding architecture that learns high-resolution, semantically meaningful, and audio-visually aligned features solely through watching videos. We show that DenseAV can discover the ``meaning'' of words and the ``location'' of sounds without explicit localization supervision. Furthermore, it automatically discovers and distinguishes between these two types of associations without supervision. We show that DenseAV's localization abilities arise from a new multi-head feature aggregation operator that directly compares dense image and audio representations for contrastive learning. In contrast, many other systems that learn ``global'' audio and video representations cannot localize words and sound. Finally, we contribute two new datasets to improve the evaluation of AV representations through speech and sound prompted semantic segmentation. On these and other datasets we show DenseAV dramatically outperforms the prior art on speech and sound prompted semantic segmentation. DenseAV outperforms the previous state-of-the-art, ImageBind, on cross-modal retrieval using fewer than half of the parameters.

A Heterotic Kähler Gravity and the Distance Conjecture
Javier José Murgas Ibarra, Paul-Konstantin Oehlmann, Fabian Ruehle, Eirik Eik Svanes
Journal of High Energy Physics, Volume 2025, article number 168 [ arXiv:2406.04393 ] published
Theoretical Physics

Abstract Deformations of the heterotic superpotential give rise to a topological holomorphic theory with similarities to both Kodaira-Spencer gravity and holomorphic Chern-Simons theory. Although the action is cubic, it is only quadratic in the complex structure deformations (the Beltrami differential). Treated separately, for large fluxes, or alternatively at large distances in the background complex structure moduli space, these fields can be integrated out to obtain a new field theory in the remaining fields, which describe the complexified hermitian and gauge degrees of freedom. We investigate properties of this new holomorphic theory, and in particular connections to the swampland distance conjecture in the context of heterotic string theory. In the process, we define a new type of symplectic cohomology theory, where the background complex structure Beltrami differential plays the role of the symplectic form.

Stochastic logic in biased coupled photonic probabilistic bits
Michael Horodynski, Charles Roques-Carmes, Yannick Salamin, Seou Choi, Jamison Sloan, Di Luo, Marin Soljačić
Communications Physics, 2025, Volume 8, Article Number 31 [ arXiv:2406.04000 ] published
Theoretical Physics

Abstract Optical computing often employs tailor-made hardware to implement specific algorithms, trading generality for improved performance in key aspects like speed and power efficiency. An important computing approach that is still missing its corresponding optical hardware is probabilistic computing, used e.g. for solving difficult combinatorial optimization problems. In this study, we propose an experimentally viable photonic approach to solve arbitrary probabilistic computing problems. Our method relies on the insight that coherent Ising machines composed of coupled and biased optical parametric oscillators can emulate stochastic logic. We demonstrate the feasibility of our approach by using numerical simulations equivalent to the full density matrix formulation of coupled optical parametric oscillators.

OccamLLM: Fast and Exact Language Model Arithmetic in a Single Step
Owen Dugan, Donato Manuel Jimenez Beneto, Charlotte Loh, Zhuo Chen, Rumen Dangovski, Marin Soljačić
[ arXiv:2406.06576 ]
Foundational AI

Abstract Despite significant advancements in text generation and reasoning, Large Language Models (LLMs) still face challenges in accurately performing complex arithmetic operations. Language model systems often enable LLMs to generate code for arithmetic operations to achieve accurate calculations. However, this approach compromises speed and security, and fine-tuning risks the language model losing prior capabilities. We propose a framework that enables exact arithmetic in a single autoregressive step, providing faster, more secure, and more interpretable LLM systems with arithmetic capabilities. We use the hidden states of a LLM to control a symbolic architecture that performs arithmetic. Our implementation using Llama 3 with OccamNet as a symbolic model (OccamLlama) achieves 100% accuracy on single arithmetic operations (+,−,×,÷,sin,cos,log,exp,√), outperforming GPT 4o with and without a code interpreter. Furthermore, OccamLlama outperforms GPT 4o with and without a code interpreter on average across a range of mathematical problem solving benchmarks, demonstrating that OccamLLMs can excel in arithmetic tasks, even surpassing much larger models. We will make our code public shortly.

SIREN: An Open Source Neutrino Injection Toolkit
Austin Schneider, Nicholas W. Kamp, Alex Y. Wen
[ arXiv:2406.01745 | code ]
Experimental Physics

Abstract Modeling of rare neutrino processes often relies on either simple approximations or expensive detector simulations. The former is often not sufficient for interactions with complex morphologies, while the latter is too time-intensive for phenomenological studies. We present SIREN (Sampling and Injection for Rare EveNts), a new tool for neutrino phenomenology and experimental searches alike that enables accurate interaction and detector geometry modeling without the overhead of detailed detector response simulations. SIREN handles the injection of rare process final states and the associated weighting calculations with the speed needed for phenomenological investigations and the detail necessary for dedicated experimental searches. The extensible design of SIREN allows it to support a wide range of experimental designs and Beyond-Standard-Model neutrino interactions. Users need only specify the physical process, detector geometry, and initial neutrino flux under consideration before they can accurately simulate a model in their detector of choice. We demonstrate the capability of SIREN through two examples: (1) Standard Model νμ deep inelastic scattering in IceCube, DUNE, and ATLAS; and (2) heavy neutral lepton interactions in MiniBooNE, MINERνA, CCM. A variety of detector geometry descriptions, interaction cross sections, and neutrino fluxes are also provided for users to get started with immediately.

Towards Universal Unfolding of Detector Effects in High-Energy Physics using Denoising Diffusion Probabilistic Models
Camila Pazos, Shuchin Aeron, Pierre-Hugues Beauchemin, Vincent Croft, Martin Klassen, Taritree Wongjirad
[ arXiv:2406.01507 ]
Experimental Physics

Abstract The unfolding of detector effects in experimental data is critical for enabling precision measurements in high-energy physics. However, traditional unfolding methods face challenges in scalability, flexibility, and dependence on simulations. We introduce a novel unfolding approach using conditional Denoising Diffusion Probabilistic Models (cDDPM). Our method utilizes the cDDPM for a non-iterative, flexible posterior sampling approach, which exhibits a strong inductive bias that allows it to generalize to unseen physics processes without explicitly assuming the underlying distribution. We test our approach by training a single cDDPM to perform multidimensional particle-wise unfolding for a variety of physics processes, including those not seen during training. Our results highlight the potential of this method as a step towards a 'universal' unfolding tool that reduces dependence on truth-level assumptions.

Exploratory Preference Optimization: Harnessing Implicit Q-Approximation for Sample-Efficient RLHF*
Tengyang Xie, Dylan J. Foster, Akshay Krishnamurthy, Corby Rosset, Ahmed Awadallah, Alexander Rakhlin
[ arXiv:2405.21046 ]
Foundational AI

Abstract Reinforcement learning from human feedback (RLHF) has emerged as a central tool for language model alignment. We consider online exploration in RLHF, which exploits interactive access to human or AI feedback by deliberately encouraging the model to produce diverse, maximally informative responses. By allowing RLHF to confidently stray from the pre-trained model, online exploration offers the possibility of novel, potentially super-human capabilities, but its full potential as a paradigm for language model training has yet to be realized, owing to computational and statistical bottlenecks in directly adapting existing reinforcement learning techniques. We propose a new algorithm for online exploration in RLHF, Exploratory Preference Optimization (XPO), which is simple and practical -- a one-line change to (online) Direct Preference Optimization (DPO; Rafailov et al., 2023) -- yet enjoys the strongest known provable guarantees and promising empirical performance. XPO augments the DPO objective with a novel and principled exploration bonus, empowering the algorithm to explore outside the support of the initial model and human feedback data. In theory, we show that XPO is provably sample-efficient and converges to a near-optimal language model policy under natural exploration conditions, irrespective of whether the initial model has good coverage. Our analysis, which builds on the observation that DPO implicitly performs a form of Q⋆-approximation (or, Bellman error minimization), combines previously disparate techniques from language modeling and theoretical reinforcement learning in a serendipitous fashion through the perspective of KL-regularized Markov decision processes. Empirically, we find that XPO is more sample-efficient than non-exploratory DPO variants in a preliminary evaluation.

QuanTA: Efficient High-Rank Fine-Tuning of LLMs with Quantum-Informed Tensor Adaptation
Zhuo Chen, Rumen Dangovski, Charlotte Loh, Owen Dugan, Di Luo, Marin Soljačić
[ arXiv:2406.00132 | code ]
Theoretical Physics

Abstract We propose Quantum-informed Tensor Adaptation (QuanTA), a novel, easy-to-implement, fine-tuning method with no inference overhead for large-scale pre-trained language models. By leveraging quantum-inspired methods derived from quantum circuit structures, QuanTA enables efficient high-rank fine-tuning, surpassing the limitations of Low-Rank Adaptation (LoRA)--low-rank approximation may fail for complicated downstream tasks. Our approach is theoretically supported by the universality theorem and the rank representation theorem to achieve efficient high-rank adaptations. Experiments demonstrate that QuanTA significantly enhances commonsense reasoning, arithmetic reasoning, and scalability compared to traditional methods. Furthermore, QuanTA shows superior performance with fewer trainable parameters compared to other approaches and can be designed to integrate with existing fine-tuning algorithms for further improvement, providing a scalable and efficient solution for fine-tuning large language models and advancing state-of-the-art in natural language processing.

Harmonic 1-forms on real loci of Calabi-Yau manifolds
Michael R. Douglas, Daniel Platt, Yidi Qi
[ arXiv:2405.19402 | code ]
Theoretical Physics

Abstract We numerically study whether there exist nowhere vanishing harmonic 1-forms on the real locus of some carefully constructed examples of Calabi-Yau manifolds, which would then give rise to potentially new examples of G2-manifolds and an explicit description of their metrics. We do this in two steps: first, we use a neural network to compute an approximate Calabi-Yau metric on each manifold. Second, we use another neural network to compute an approximately harmonic 1-form with respect to the approximate metric, and then inspect the found solution. On two manifolds existence of a nowhere vanishing harmonic 1-form can be ruled out using differential geometry. The real locus of a third manifold is diffeomorphic to S1×S2, and our numerics suggest that when the Calabi-Yau metric is close to a singular limit, then it admits a nowhere vanishing harmonic 1-form. We explain how such an approximate solution could potentially be used in a numerically verified proof for the fact that our example manifold must admit a nowhere vanishing harmonic 1-form.

Survival of the Fittest Representation: A Case Study with Modular Addition
Xiaoman Delores Ding, Zifan Carl Guo, Eric J. Michaud, Ziming Liu, Max Tegmark
[ arXiv:2405.17420 | code ]
Foundational AI

Abstract When a neural network can learn multiple distinct algorithms to solve a task, how does it 'choose' between them during training? To approach this question, we take inspiration from ecology: when multiple species coexist, they eventually reach an equilibrium where some survive while others die out. Analogously, we suggest that a neural network at initialization contains many solutions (representations and algorithms), which compete with each other under pressure from resource constraints, with the 'fittest' ultimately prevailing. To investigate this Survival of the Fittest hypothesis, we conduct a case study on neural networks performing modular addition, and find that these networks' multiple circular representations at different Fourier frequencies undergo such competitive dynamics, with only a few circles surviving at the end. We find that the frequencies with high initial signals and gradients, the 'fittest,' are more likely to survive. By increasing the embedding dimension, we also observe more surviving frequencies. Inspired by the Lotka-Volterra equations describing the dynamics between species, we find that the dynamics of the circles can be nicely characterized by a set of linear differential equations. Our results with modular addition show that it is possible to decompose complicated representations into simpler components, along with their basic interactions, to offer insight on the training dynamics of representations.

From Neurons to Neutrons: A Case Study in Interpretability
Ouail Kitouni, Niklas Nolte, Víctor Samuel Pérez-Díaz, Sokratis Trifinopoulos, Mike Williams
[ arXiv:2405.17425 | code ]
Experimental Physics

Abstract Mechanistic Interpretability (MI) promises a path toward fully understanding how neural networks make their predictions. Prior work demonstrates that even when trained to perform simple arithmetic, models can implement a variety of algorithms (sometimes concurrently) depending on initialization and hyperparameters. Does this mean neuron-level interpretability techniques have limited applicability? We argue that high-dimensional neural networks can learn low-dimensional representations of their training data that are useful beyond simply making good predictions. Such representations can be understood through the mechanistic interpretability lens and provide insights that are surprisingly faithful to human-derived domain knowledge. This indicates that such approaches to interpretability can be useful for deriving a new understanding of a problem from models trained to solve it. As a case study, we extract nuclear physics concepts by studying models trained to reproduce nuclear data.

How Do Transformers ‘Do’ Physics? Investigating the Simple Harmonic Oscillator
Subhash Kantamneni, Ziming Liu, Max Tegmark
Entropy, Volume 26, Issue 11, 2024 [ arXiv:2405.17209 | code ] published
Foundational AI

Abstract How do transformers model physics? Do transformers model systems with interpretable analytical solutions, or do they create 'alien physics' that are difficult for humans to decipher? We take a step in demystifying this larger puzzle by investigating the simple harmonic oscillator (SHO), ẍ+2γẋ+ω20x=0, one of the most fundamental systems in physics. Our goal is to identify the methods transformers use to model the SHO, and to do so we hypothesize and evaluate possible methods by analyzing the encoding of these methods' intermediates. We develop four criteria for the use of a method within the simple testbed of linear regression, where our method is y=wx and our intermediate is w - (1) Can the intermediate be predicted from hidden states? (2) Is the intermediate's encoding quality correlated with model performance? (3) Can the majority of variance in hidden states be explained by the intermediate? (4) Can we intervene on hidden states to produce predictable outcomes? Armed with these two correlational (1,2), weak causal (3) and strong causal (4) criteria, we determine that transformers use known numerical methods to model trajectories of the simple harmonic oscillator, specifically the matrix exponential method. Our analysis framework can conveniently extend to high-dimensional linear systems and nonlinear systems, which we hope will help reveal the 'world model' hidden in transformers.

Improved Distribution Matching Distillation for Fast Image Synthesis
Tianwei Yin, Michaël Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Fredo Durand, William T. Freeman
[ arXiv:2405.14867 | code ]
Foundational AI

Abstract Recent approaches have shown promises distilling diffusion models into efficient one-step generators. Among them, Distribution Matching Distillation (DMD) produces one-step generators that match their teacher in distribution, without enforcing a one-to-one correspondence with the sampling trajectories of their teachers. However, to ensure stable training, DMD requires an additional regression loss computed using a large set of noise-image pairs generated by the teacher with many steps of a deterministic sampler. This is costly for large-scale text-to-image synthesis and limits the student's quality, tying it too closely to the teacher's original sampling paths. We introduce DMD2, a set of techniques that lift this limitation and improve DMD training. First, we eliminate the regression loss and the need for expensive dataset construction. We show that the resulting instability is due to the fake critic not estimating the distribution of generated samples accurately and propose a two time-scale update rule as a remedy. Second, we integrate a GAN loss into the distillation procedure, discriminating between generated samples and real images. This lets us train the student model on real data, mitigating the imperfect real score estimation from the teacher model, and enhancing quality. Lastly, we modify the training procedure to enable multi-step sampling. We identify and address the training-inference input mismatch problem in this setting, by simulating inference-time generator samples during training time. Taken together, our improvements set new benchmarks in one-step image generation, with FID scores of 1.28 on ImageNet-64x64 and 8.35 on zero-shot COCO 2014, surpassing the original teacher despite a 500X reduction in inference cost. Further, we show our approach can generate megapixel images by distilling SDXL, demonstrating exceptional visual quality among few-step methods.

Not All Language Model Features Are Linear
Joshua Engels, Isaac Liao, Eric J. Michaud, Wes Gurnee, Max Tegmark
ICLR 2025 [ arXiv:2405.14860 | code ] published
Foundational AI

Abstract Recent work has proposed the linear representation hypothesis: that language models perform computation by manipulating one-dimensional representations of concepts ('features') in activation space. In contrast, we explore whether some language model representations may be inherently multi-dimensional. We begin by developing a rigorous definition of irreducible multi-dimensional features based on whether they can be decomposed into either independent or non-co-occurring lower-dimensional features. Motivated by these definitions, we design a scalable method that uses sparse autoencoders to automatically find multi-dimensional features in GPT-2 and Mistral 7B. These auto-discovered features include strikingly interpretable examples, e.g. circular features representing days of the week and months of the year. We identify tasks where these exact circles are used to solve computational problems involving modular arithmetic in days of the week and months of the year. Finally, we provide evidence that these circular features are indeed the fundamental unit of computation in these tasks with intervention experiments on Mistral 7B and Llama 3 8B, and we find further circular representations by breaking down the hidden states for these tasks into interpretable components.

Lorentz-Equivariant Geometric Algebra Transformers for High-Energy Physics
Jonas Spinner, Victor Bresó, Pim de Haan, Tilman Plehn, Jesse Thaler, Johann Brehmer
NeurIPS 2024 [ arXiv:2405.14806 | code ] published
Experimental Physics

Abstract Extracting scientific understanding from particle-physics experiments requires solving diverse learning problems with high precision and good data efficiency. We propose the Lorentz Geometric Algebra Transformer (L-GATr), a new multi-purpose architecture for high-energy physics. L-GATr represents high-energy data in a geometric algebra over four-dimensional space-time and is equivariant under Lorentz transformations, the symmetry group of relativistic kinematics. At the same time, the architecture is a Transformer, which makes it versatile and scalable to large systems. L-GATr is first demonstrated on regression and classification tasks from particle physics. We then construct the first Lorentz-equivariant generative model: a continuous normalizing flow based on an L-GATr network, trained with Riemannian flow matching. Across our experiments, L-GATr is on par with or outperforms strong domain-specific baselines.

Multistable Shape from Shading Emerges from Patch Diffusion
Xinran Nicole Han, Todd Zickler, Ko Nishino
NeurIPS 2024 [ arXiv:2405.14530 ] published
Foundational AI

Abstract Models for monocular shape reconstruction of surfaces with diffuse reflection -- shape from shading -- ought to produce distributions of outputs, because there are fundamental mathematical ambiguities of both continuous (e.g., bas-relief) and discrete (e.g., convex/concave) varieties which are also experienced by humans. Yet, the outputs of current models are limited to point estimates or tight distributions around single modes, which prevent them from capturing these effects. We introduce a model that reconstructs a multimodal distribution of shapes from a single shading image, which aligns with the human experience of multistable perception. We train a small denoising diffusion process to generate surface normal fields from 16×16 patches of synthetic images of everyday 3D objects. We deploy this model patch-wise at multiple scales, with guidance from inter-patch shape consistency constraints. Despite its relatively small parameter count and predominantly bottom-up structure, we show that multistable shape explanations emerge from this model for 'ambiguous' test images that humans experience as being multistable. At the same time, the model produces veridical shape estimates for object-like images that include distinctive occluding contours and appear less ambiguous. This may inspire new architectures for stochastic 3D shape perception that are more efficient and better aligned with human experience.

Unsupervised Searches for Cosmological Parity Violation: Improving Detection Power with the Neural Field Scattering Transform
Matthew Craigie, Peter L. Taylor, Yuan-Sen Ting, Carolina Cuesta-Lazaro, Rossana Ruggeri, Tamara M. Davis
[ arXiv:2405.13083 ]
Astrophysics

Abstract Recent studies using four-point correlations suggest a parity violation in the galaxy distribution, though the significance of these detections is sensitive to the choice of simulation used to model the noise properties of the galaxy distribution. In a recent paper, we introduce an unsupervised learning approach which offers an alternative method that avoids the dependence on mock catalogs, by learning parity violation directly from observational data. However, the Convolutional Neural Network (CNN) model utilized by our previous unsupervised approach struggles to extend to more realistic scenarios where data is limited. We propose a novel method, the Neural Field Scattering Transform (NFST), which enhances the Wavelet Scattering Transform (WST) technique by adding trainable filters, parameterized as a neural field. We first tune the NFST model to detect parity violation in a simplified dataset, then compare its performance against WST and CNN benchmarks across varied training set sizes. We find the NFST can detect parity violation with 4× less data than the CNN and 32× less than the WST. Furthermore, in cases with limited data the NFST can detect parity violation with up to 6σ confidence, where the WST and CNN fail to make any detection. We identify that the added flexibility of the NFST, and particularly the ability to learn asymmetric filters, as well as the specific symmetries built into the NFST architecture, contribute to its improved performance over the benchmark models. We further demonstrate that the NFST is readily interpretable, which is valuable for physical applications such as the detection of parity violation.

Resonant Neutrino Flavor Conversion in the Atmosphere
Connor Sponsler, Matheus Hostert, Ivan Martinez-Soler, Carlos A. Argüelles
[ arXiv:2405.12140 ]
Experimental Physics

Abstract Neutrinos produced in the atmosphere traverse a column density of air before being detected at neutrino observatories like IceCube or KM3NeT. In this work, we extend the neutrino flavor evolution in the {nuSQuIDS} code accounting for the varying height of neutrino production and the variable air density in the atmosphere. These effects can lead to sizeable spectral distortions in standard neutrino oscillations and are crucial to accurately describe some new physics scenarios. As an example, we study a model of quasi-sterile neutrinos that induce resonant flavor conversions at neutrino energies of O(300) MeV in matter densities of 1 g/cm3. In atmospheric air densities, the same resonance is then realized at neutrino energies of O(300−700)~GeV. We find that the new resonance can deplete the $\nu$μ+$\bar{\nu}$μ flux at the IceCube Neutrino Observatory by as much as 10% in the direction of the horizon.

Diffusion-HMC: Parameter Inference with Diffusion Model driven Hamiltonian Monte Carlo
Nayantara Mudur, Carolina Cuesta-Lazaro, Douglas P. Finkbeiner
The Astrophysical Journal, 978, 64, 2025 [ arXiv:2405.05255 | code ] published
Astrophysics

Abstract Diffusion generative models have excelled at diverse image generation and reconstruction tasks across fields. A less explored avenue is their application to discriminative tasks involving regression or classification problems. The cornerstone of modern cosmology is the ability to generate predictions for observed astrophysical fields from theory and constrain physical models from observations using these predictions. This work uses a single diffusion generative model to address these interlinked objectives -- as a surrogate model or emulator for cold dark matter density fields conditional on input cosmological parameters, and as a parameter inference model that solves the inverse problem of constraining the cosmological parameters of an input field. The model is able to emulate fields with summary statistics consistent with those of the simulated target distribution. We then leverage the approximate likelihood of the diffusion generative model to derive tight constraints on cosmology by using the Hamiltonian Monte Carlo method to sample the posterior on cosmological parameters for a given test image. Finally, we demonstrate that this parameter inference approach is more robust to the addition of noise than baseline parameter inference networks.

OptPDE: Discovering Novel Integrable Systems via AI-Human Collaboration
Subhash Kantamneni, Ziming Liu, Max Tegmark
[ arXiv:2405.04484 ]
Foundational AI

Abstract Integrable partial differential equation (PDE) systems are of great interest in natural science, but are exceedingly rare and difficult to discover. To solve this, we introduce OptPDE, a first-of-its-kind machine learning approach that Optimizes PDEs' coefficients to maximize their number of conserved quantities, nCQ, and thus discover new integrable systems. We discover four families of integrable PDEs, one of which was previously known, and three of which have at least one conserved quantity but are new to the literature to the best of our knowledge. We investigate more deeply the properties of one of these novel PDE families, ut=(ux+a2uxxx)3. Our paper offers a promising schema of AI-human collaboration for integrable system discovery; machine learning generates interpretable hypotheses for possible integrable systems, which human scientists can verify and analyze, to truly close the discovery loop.

Bounds and Dualities of Type II Little String Theories
Florent Baume, Paul-Konstantin Oehlmann, Fabian Ruehle
Journal of High Energy Physics, Volume 2024, article number 149, 2024 [ arXiv:2405.03877 ] published
Theoretical Physics

Abstract We explore the symmetry structure of Type II Little String Theories and their T-dualities. We construct these theories both from the bottom-up perspective starting with seed Superconformal Field Theories, and from the top-down using F-/M-theory. By exploiting anomaly inflow and unitarity of the LST worldsheet theory, we derive strong conditions on the possible 6D bulk theories and their flavor algebras. These constraints continue to apply if gravity is coupled to the theory. We also study the higher form symmetry structure of these theories and show how they get exchanged under T-duality. Finally, we comment on seemingly consistent bottom-up Little String Theories that cannot be constructed from the top-down approach.

Measurement of atmospheric neutrino oscillation parameters using convolutional neural networks with 9.3 years of data in IceCube DeepCore
IceCube Collaboration
Physical Review Letters, 134, 091801, 2025 [ arXiv:2405.02163 ] published
Experimental Physics

Abstract The DeepCore sub-detector of the IceCube Neutrino Observatory provides access to neutrinos with energies above approximately 5 GeV. Data taken between 2012-2021 (3,387 days) are utilized for an atmospheric νμ disappearance analysis that studied 150,257 neutrino-candidate events with reconstructed energies between 5-100 GeV. An advanced reconstruction based on a convolutional neural network is applied, providing increased signal efficiency and background suppression, resulting in a measurement with both significantly increased statistics compared to previous DeepCore oscillation results and high neutrino purity. For the normal neutrino mass ordering, the atmospheric neutrino oscillation parameters and their 1σ errors are measured to be Δm232 = 2.40+0.05−0.04×10−3 eV2 and sin2θ23=0.54+0.04−0.03. The results are the most precise to date using atmospheric neutrinos, and are compatible with measurements from other neutrino detectors including long-baseline accelerator experiments.

A Parameter-Masked Mock Data Challenge for Beyond-Two-Point Galaxy Clustering Statistics
Beyond-2pt Collaboration - Elisabeth Krause, Yosuke Kobayashi, Andrés N. Salcedo, Mikhail M. Ivanov, Tom Abel, Kazuyuki Akitsu, Raul E. Angulo, Giovanni Cabass, Sofia Contarini, Carolina Cuesta-Lazaro, ChangHoon Hahn, Nico Hamaus, Donghui Jeong, Chirag Modi, Nhat-Minh Nguyen, Takahiro Nishimichi, Enrique Paillas, Marcos Pellejero Ibañez, Oliver H. E. Philcox, Alice Pisani, Fabian Schmidt, Satoshi Tanaka, Giovanni Verza, Sihan Yuan, Matteo Zennaro
[ arXiv:2405.02252 ]
Astrophysics

Abstract The last few years have seen the emergence of a wide array of novel techniques for analyzing high-precision data from upcoming galaxy surveys, which aim to extend the statistical analysis of galaxy clustering data beyond the linear regime and the canonical two-point (2pt) statistics. We test and benchmark some of these new techniques in a community data challenge 'Beyond-2pt', initiated during the Aspen 2022 Summer Program 'Large-Scale Structure Cosmology beyond 2-Point Statistics,' whose first round of results we present here. The challenge dataset consists of high-precision mock galaxy catalogs for clustering in real space, redshift space, and on a light cone. Participants in the challenge have developed end-to-end pipelines to analyze mock catalogs and extract unknown ('masked') cosmological parameters of the underlying ΛCDM models with their methods. The methods represented are density-split clustering, nearest neighbor statistics, BACCO power spectrum emulator, void statistics, LEFTfield field-level inference using effective field theory (EFT), and joint power spectrum and bispectrum analyses using both EFT and simulation-based inference. In this work, we review the results of the challenge, focusing on problems solved, lessons learned, and future research needed to perfect the emerging beyond-2pt approaches. The unbiased parameter recovery demonstrated in this challenge by multiple statistics and the associated modeling and inference frameworks supports the credibility of cosmology constraints from these methods. The challenge data set is publicly available and we welcome future submissions from methods that are not yet represented.

Multi-filter UV to NIR Data-driven Light Curve Templates for Stripped Envelope Supernovae
Somayeh Khakpash, Federica B. Bianco, Maryam Modjaz, Willow F. Fortino, Alexander Gagliano, Conor Larison, Tyler A. Pritchard
The Astrophysical Journal, Volume 275, Number 2, 2024 [ arXiv:2405.01672 ] published
Astrophysics

Abstract While the spectroscopic classification scheme for Stripped envelope supernovae (SESNe) is clear, and we know that they originate from massive stars that lost some or all their envelopes of Hydrogen and Helium, the photometric evolution of classes within this family is not fully characterized. Photometric surveys, like the Vera C. Rubin Legacy Survey of Space and Time, will discover tens of thousands of transients each night and spectroscopic follow-up will be limited, prompting the need for photometric classification and inference based solely on photometry. We have generated 54 data-driven photometric templates for SESNe of subtypes IIb, Ib, Ic, Ic-bl, and Ibn in U/u, B, g, V, R/r, I/i, J, H, Ks, and Swift w2, m2, w1 bands using Gaussian Processes and a multi-survey dataset composed of all well-sampled open-access light curves (165 SESNe, 29531 data points) from the Open Supernova Catalog. We use our new templates to assess the photometric diversity of SESNe by comparing final per-band subtype templates with each other and with individual, unusual and prototypical SESNe. We find that SNe Ibns and Ic-bl exhibit a distinctly faster rise and decline compared to other subtypes. We also evaluate the behavior of SESNe in the PLAsTiCC and ELAsTiCC simulations of LSST light curves highlighting differences that can bias photometric classification models trained on the simulated light curves. Finally, we investigate in detail the behavior of fast-evolving SESNe (including SNe Ibn) and the implications of the frequently observed presence of two peaks in their light curves.

Learning Force Control for Legged Manipulation
Tifanny Portela, Gabriel B. Margolis, Yandong Ji, Pulkit Agrawal
[ arXiv:2405.01402 | code ]
Foundational AI

Abstract Controlling contact forces during interactions is critical for locomotion and manipulation tasks. While sim-to-real reinforcement learning (RL) has succeeded in many contact-rich problems, current RL methods achieve forceful interactions implicitly without explicitly regulating forces. We propose a method for training RL policies for direct force control without requiring access to force sensing. We showcase our method on a whole-body control platform of a quadruped robot with an arm. Such force control enables us to perform gravity compensation and impedance control, unlocking compliant whole-body manipulation. The learned whole-body controller with variable compliance makes it intuitive for humans to teleoperate the robot by only commanding the manipulator, and the robot's body adjusts automatically to achieve the desired position and force. Consequently, a human teleoperator can easily demonstrate a wide variety of loco-manipulation tasks. To the best of our knowledge, we provide the first deployment of learned whole-body force control in legged manipulators, paving the way for more versatile and adaptable legged robots.

KAN: Kolmogorov-Arnold Networks
Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Soljačić, Thomas Y. Hou, Max Tegmark
[ arXiv:2404.19756 | code ]
Foundational AI

Abstract Inspired by the Kolmogorov-Arnold representation theorem, we propose Kolmogorov-Arnold Networks (KANs) as promising alternatives to Multi-Layer Perceptrons (MLPs). While MLPs have fixed activation functions on nodes ("neurons"), KANs have learnable activation functions on edges ("weights"). KANs have no linear weights at all -- every weight parameter is replaced by a univariate function parametrized as a spline. We show that this seemingly simple change makes KANs outperform MLPs in terms of accuracy and interpretability. For accuracy, much smaller KANs can achieve comparable or better accuracy than much larger MLPs in data fitting and PDE solving. Theoretically and empirically, KANs possess faster neural scaling laws than MLPs. For interpretability, KANs can be intuitively visualized and can easily interact with human users. Through two examples in mathematics and physics, KANs are shown to be useful collaborators helping scientists (re)discover mathematical and physical laws. In summary, KANs are promising alternatives for MLPs, opening opportunities for further improving todays deep learning models which rely heavily on MLPs.

SN 2024ggi in NGC 3621: Rising Ionization in a Nearby, CSM-Interacting Type II Supernova
CW. V. Jacobson-Galán, K. W. Davis, C. D. Kilpatrick, L. Dessart, R. Margutti, R. Chornock, R. J. Foley, P. Arunachalam, K. Auchettl, C. R. Bom, R. Cartier, D. A. Coulter, G. Dimitriadis, D. Dickinson, M. R. Drout, A. T. Gagliano, C. Gall, B. Garretson, L. Izzo, D. O. Jones, N. LeBaron, H.-Y. Miao, D. Milisavljevic, Y.-C. Pan, A. Rest, C. Rojas-Bravo, A. Santos, H. Sears, B. M. Subrayan, K. Taggart, S. Tinyanont
The Astrophysical Journal, 2024, Volume 972, Number 2 [ arXiv:2404.19006 ] published
Astrophysics

Abstract We present UV/optical/NIR observations and modeling of supernova (SN) 2024ggi, a type II supernova (SN II) located in NGC 3621 at 7.2 Mpc. Early-time ('flash') spectroscopy of SN 2024ggi within +0.8 days of discovery shows emission lines of H I, He I, C III, and N III with a narrow core and broad, symmetric wings (i.e., IIn-like) arising from the photoionized, optically-thick, unshocked circumstellar material (CSM) that surrounded the progenitor star at shock breakout. By the next spectral epoch at +1.5 days, SN 2024ggi showed a rise in ionization as emission lines of He II, C IV, N IV/V and O V became visible. This phenomenon is temporally consistent with a blueward shift in the UV/optical colors, both likely the result of shock breakout in an extended, dense CSM. The IIn-like features in SN 2024ggi persist on a timescale of tIIn=3.8±1.6 days at which time a reduction in CSM density allows the detection of Doppler broadened features from the fastest SN material. SN 2024ggi has peak UV/optical absolute magnitudes of Mw2=−18.7 mag and Mg=−18.1 mag that are consistent with the known population of CSM-interacting SNe II. Comparison of SN 2024ggi with a grid of radiation hydrodynamics and non-local thermodynamic equilibrium (nLTE) radiative-transfer simulations suggests a progenitor mass-loss rate of M˙=10−2M⊙ yr−1 (vw = 50 km/s), confined to a distance of r<5×1014 cm. Assuming a wind velocity of vw = 50 km/s, the progenitor star underwent an enhanced mass-loss episode in the last ~3 years before explosion.

Position-space renormalization schemes for four-quark operators in HQET
Joshua Lin, William Detmold, Stefan Meinel
Journal of High Energy Physics, Volume 2024, article number 188 [ arXiv:2404.16191 ] published
Theoretical Physics

Abstract X-space schemes are gauge-invariant, regulator-independent renormalization schemes that are defined by requiring position-space correlation functions of gauge invariant operators to be equal to their noninteracting values at particular kinematic points. These schemes can be used to nonperturbatively renormalize composite operators in Lattice Quantum Chromodynamics (LQCD), and by computing matching coefficients between the X-space scheme and MSbar in the dimensionally-regulated continuum, matrix elements calculated with LQCD can be converted to MSbar-renormalized matrix elements. Using X-space schemes for Heavy Quark Effective Theory (HQET) operators has the additional benefit that appropriate ratios of position-space correlation functions cancel the power divergent static-quark self-energy of Lattice HQET nonperturbatively. This work presents the O(αS) matching coefficients between X-space renormalized four-quark flavor-nonsinglet HQET operators relevant for the lifetimes of charm- and bottom-hadrons, and four-quark HQET operators relevant for mixing between neutral mesons containing a heavy quark, such as B-Bbar mixing.

Learning Galaxy Intrinsic Alignment Correlations
Sneh Pandya, Yuanyuan Yang, Nicholas Van Alfen, Jonathan Blazek, Robin Walters
[ arXiv:2404.13702 ]
Astrophysics

Abstract The intrinsic alignments (IA) of galaxies, regarded as a contaminant in weak lensing analyses, represents the correlation of galaxy shapes due to gravitational tidal interactions and galaxy formation processes. As such, understanding IA is paramount for accurate cosmological inferences from weak lensing surveys; however, one limitation to our understanding and mitigation of IA is expensive simulation-based modeling. In this work, we present a deep learning approach to emulate galaxy position-position (ξ), position-orientation (ω), and orientation-orientation (η) correlation function measurements and uncertainties from halo occupation distribution-based mock galaxy catalogs. We find strong Pearson correlation values with the model across all three correlation functions and further predict aleatoric uncertainties through a mean-variance estimation training procedure. ξ(r) predictions are generally accurate to ≤10%. Our model also successfully captures the underlying signal of the noisier correlations ω(r) and η(r), although with a lower average accuracy. We find that the model performance is inhibited by the stochasticity of the data, and will benefit from correlations averaged over multiple data realizations. Our code will be made open source upon journal publication.

PhysDreamer: Physics-Based Interaction with 3D Objects via Video Generation
Tianyuan Zhang, Hong-Xing Yu, Rundi Wu, Brandon Y. Feng, Changxi Zheng, Noah Snavely, Jiajun Wu, William T. Freeman
Springer Nature Link, 2024 [ arXiv:2404.13026 | code ] published
Foundational AI

Abstract Realistic object interactions are crucial for creating immersive virtual experiences, yet synthesizing realistic 3D object dynamics in response to novel interactions remains a significant challenge. Unlike unconditional or text-conditioned dynamics generation, action-conditioned dynamics requires perceiving the physical material properties of objects and grounding the 3D motion prediction on these properties, such as object stiffness. However, estimating physical material properties is an open problem due to the lack of material ground-truth data, as measuring these properties for real objects is highly difficult. We present PhysDreamer, a physics-based approach that endows static 3D objects with interactive dynamics by leveraging the object dynamics priors learned by video generation models. By distilling these priors, PhysDreamer enables the synthesis of realistic object responses to novel interactions, such as external forces or agent manipulations. We demonstrate our approach on diverse examples of elastic objects and evaluate the realism of the synthesized interactions through a user study. PhysDreamer takes a step towards more engaging and realistic virtual experiences by enabling static 3D objects to dynamically respond to interactive stimuli in a physically plausible manner.

Practical applications of machine-learned flows on gauge fields
Ryan Abbott, Michael S. Albergo, Denis Boyda, Daniel C. Hackett, Gurtej Kanwar, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2404.11674 ]
Theoretical Physics

Abstract Normalizing flows are machine-learned maps between different lattice theories which can be used as components in exact sampling and inference schemes. Ongoing work yields increasingly expressive flows on gauge fields, but it remains an open question how flows can improve lattice QCD at state-of-the-art scales. We discuss and demonstrate two applications of flows in replica exchange (parallel tempering) sampling, aimed at improving topological mixing, which are viable with iterative improvements upon presently available flows.

Foundations of Computer Vision
Antonio Torralba, Phillip Isola, William T. Freeman
MIT Press (Book), 2024 [ ] published
Foundational AI

Abstract An accessible, authoritative, and up-to-date computer vision textbook offering a comprehensive introduction to the foundations of the field that incorporates the latest deep learning advances.

Machine learning has revolutionized computer vision, but the methods of today have deep roots in the history of the field. Providing a much-needed modern treatment, this accessible and up-to-date textbook comprehensively introduces the foundations of computer vision while incorporating the latest deep learning advances. Taking a holistic approach that goes beyond machine learning, it addresses fundamental issues in the task of vision and the relationship of machine vision to human perception. Foundations of Computer Vision covers topics not standard in other texts, including transformers, diffusion models, statistical image models, issues of fairness and ethics, and the research process. To emphasize intuitive learning, concepts are presented in short, lucid chapters alongside extensive illustrations, questions, and examples. Written by leaders in the field and honed by a decade of classroom experience, this engaging and highly teachable book offers an essential next-generation view of computer vision.

Multiscale Normalizing Flows for Gauge Theories
Ryan Abbott, Michael S. Albergo, Denis Boyda, Daniel C. Hackett, Gurtej Kanwar, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2404.10819 ]
Theoretical Physics

Abstract Scale separation is an important physical principle that has previously enabled algorithmic advances such as multigrid solvers. Previous work on normalizing flows has been able to utilize scale separation in the context of scalar field theories, but the principle has been largely unexploited in the context of gauge theories. This work gives an overview of a new method for generating gauge fields using hierarchical normalizing flow models. This method builds gauge fields from the outside in, allowing different parts of the model to focus on different scales of the problem. Numerical results are presented for U(1) and SU(3) gauge theories in 2, 3, and 4 spacetime dimensions.

TENG: Time-Evolving Natural Gradient for Solving PDEs with Deep Neural Net
Zhuo Chen, Jacob McCarran, Esteban Vizcaino, Marin Soljačić, Di Luo
Open Review, Submission Number 10038 [ arXiv:2404.10771 | code ] published
Theoretical Physics

Abstract Partial differential equations (PDEs) are instrumental for modeling dynamical systems in science and engineering. The advent of neural networks has initiated a significant shift in tackling these complexities though challenges in accuracy persist, especially for initial value problems. In this paper, we introduce the Time-Evolving Natural Gradient (TENG), generalizing time-dependent variational principles and optimization-based time integration, leveraging natural gradient optimization to obtain high accuracy in neural-network-based PDE solutions. Our comprehensive development includes algorithms like TENG-Euler and its high-order variants, such as TENG-Heun, tailored for enhanced precision and efficiency. TENG's effectiveness is further validated through its performance, surpassing current leading methods and achieving machine precision in step-by-step optimizations across a spectrum of PDEs, including the heat equation, Allen-Cahn equation, and Burgers' equation.

The Frozen Phase of Heterotic F-theory Duality
Paul-Konstantin Oehlmann, Fabian Ruehle, Benjamin Sung
Journal of High Energy Physics, Volume 2024, Article Number 295 [ arXiv:2404.02191 ] published
Theoretical Physics

Abstract We study the duality between the Spin(32)/ℤ2 heterotic string without vector structure and F-theory with frozen singularities. We give a complete description in theories with 6d =(1,0) supersymmetry and identify the duals of Spin(32)/ℤ2-instantons on ADE singularities without vector structure in the frozen phase of F-theory using an ansatz introduced by Bhardwaj, Morrison, Tachikawa, and Tomasiello. As a consequence, we obtain a strongly coupled description of orbifold phases of type I string theory without vector structure, substantially expanding the list of known examples of 6d F-theory compactifications with frozen singularities. Supergravity theories can be fused from these instanton theories, in a way that commutes with switching off vector structure, which we use to propose new consistency checks via neutral hypermultiplet counting. Finally, we describe various Higgsings of this duality, and comment on constraints on higher form symmetries.

Anomaly Detection and Approximate Similarity Searches of Transients in Real-time Data Streams
P. D. Aleo, A. W. Engel, G. Narayan, C. R. Angus, K. Malanchev, K. Auchettl, V. F. Baldassare, A. Berres, T. J. L. de Boer, B. M. Boyd, K. C. Chambers, K. W. Davis, N. Esquivel, D. Farias, R. J. Foley, A. Gagliano, C. Gall, H. Gao, S. Gomez, M. Grayling, D. O. Jones, C.-C. Lin, E. A. Magnier, K. S. Mandel, T. Matheson, S. I. Raimundo, V. G. Shah, M. D. Soraisam, K. M. de Soto, S. Vicencio, V. A. Villar, R. J. Wainscoat
The Astrophysical Journal, Volume 974, Number 2, 2024 [ arXiv:2404.01235 ] published
Astrophysics

Abstract We present LAISS (Lightcurve Anomaly Identification and Similarity Search), an automated pipeline to detect anomalous astrophysical transients in real-time data streams. We deploy our anomaly detection model on the nightly ZTF Alert Stream via the ANTARES broker, identifying a manageable ∼1-5 candidates per night for expert vetting and coordinating follow-up observations. Our method leverages statistical light-curve and contextual host-galaxy features within a random forest classifier, tagging transients of rare classes (spectroscopic anomalies), of uncommon host-galaxy environments (contextual anomalies), and of peculiar or interaction-powered phenomena (behavioral anomalies). Moreover, we demonstrate the power of a low-latency (∼ms) approximate similarity search method to find transient analogs with similar light-curve evolution and host-galaxy environments. We use analogs for data-driven discovery, characterization, (re-)classification, and imputation in retrospective and real-time searches. To date we have identified ∼50 previously known and previously missed rare transients from real-time and retrospective searches, including but not limited to: SLSNe, TDEs, SNe IIn, SNe IIb, SNe Ia-CSM, SNe Ia-91bg-like, SNe Ib, SNe Ic, SNe Ic-BL, and M31 novae. Lastly, we report the discovery of 325 total transients, all observed between 2018-2021 and absent from public catalogs (∼1% of all ZTF Astronomical Transient reports to the Transient Name Server through 2021). These methods enable a systematic approach to finding the "needle in the haystack" in large-volume data streams. Because of its integration with the ANTARES broker, LAISS is built to detect exciting transients in Rubin data.

Feature Splatting: Language-Driven Physics-Based Scene Synthesis and Editing
Ri-Zhao Qiu, Ge Yang, Weijia Zeng, Xiaolong Wang
Springer Nature Link, 2024 [ arXiv:2404.01223 | code ] published
Foundational AI

Abstract Scene representations using 3D Gaussian primitives have produced excellent results in modeling the appearance of static and dynamic 3D scenes. Many graphics applications, however, demand the ability to manipulate both the appearance and the physical properties of objects. We introduce Feature Splatting, an approach that unifies physics-based dynamic scene synthesis with rich semantics from vision language foundation models that are grounded by natural language. Our first contribution is a way to distill high-quality, object-centric vision-language features into 3D Gaussians, that enables semi-automatic scene decomposition using text queries. Our second contribution is a way to synthesize physics-based dynamics from an otherwise static scene using a particle-based simulator, in which material properties are assigned automatically via text queries. We ablate key techniques used in this pipeline, to illustrate the challenge and opportunities in using feature-carrying 3D Gaussians as a unified format for appearance, geometry, material properties and semantics grounded on natural language. Project website: https://feature-splatting.github.io/

Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data
Matthias Gerstgrasser, Rylan Schaeffer, Apratim Dey, Rafael Rafailov, Henry Sleight, John Hughes, Tomasz Korbak, Rajashree Agrawal, Dhruv Pai, Andrey Gromov, Daniel A. Roberts, Diyi Yang, David L. Donoho, Sanmi Koyejo
[ arXiv:2404.01413 ]
Foundational AI

Abstract The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops discovered that such loops can lead to model collapse, a phenomenon where performance progressively degrades with each model-fitting iteration until the latest model becomes useless. However, several recent papers studying model collapse assumed that new data replace old data over time rather than assuming data accumulate over time. In this paper, we compare these two settings and show that accumulating data prevents model collapse. We begin by studying an analytically tractable setup in which a sequence of linear models are fit to the previous models predictions. Previous work showed if data are replaced, the test error increases linearly with the number of model-fitting iterations; we extend this result by proving that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations. We next empirically test whether accumulating data similarly prevents model collapse by pretraining sequences of language models on text corpora. We confirm that replacing data does indeed cause model collapse, then demonstrate that accumulating data prevents model collapse; these results hold across a range of model sizes, architectures and hyperparameters. We further show that similar results hold for other deep generative models on real data: diffusion models for molecule generation and variational autoencoders for image generation. Our work provides consistent theoretical and empirical evidence that data accumulation mitigates model collapse.

Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
Samuel Marks, Can Rager, Eric J. Michaud, Yonatan Belinkov, David Bau, Aaron Mueller
ICLR 2025 [ arXiv:2403.19647 | code ] published
Foundational AI

Abstract We introduce methods for discovering and applying sparse feature circuits. These are causally implicated subnetworks of human-interpretable features for explaining language model behaviors. Circuits identified in prior work consist of polysemantic and difficult-to-interpret units like attention heads or neurons, rendering them unsuitable for many downstream applications. In contrast, sparse feature circuits enable detailed understanding of unanticipated mechanisms. Because they are based on fine-grained units, sparse feature circuits are useful for downstream tasks: We introduce SHIFT, where we improve the generalization of a classifier by ablating features that a human judges to be task-irrelevant. Finally, we demonstrate an entirely unsupervised and scalable interpretability pipeline by discovering thousands of sparse feature circuits for automatically discovered model behaviors.

A machine-learning pipeline for real-time detection of gravitational waves from compact binary coalescences
Ethan Marx, William Benoit, Alec Gunny, Rafia Omer, Deep Chatterjee, Ricco C. Venterea, Lauren Wills, Muhammed Saleem, Eric Moreno, Ryan Raikman, Ekaterina Govorkova, Dylan Rankin, Michael W. Coughlin, Philip Harris, Erik Katsavounidis
[ arXiv:2403.18661 | code ]
Astrophysics

Abstract The promise of multi-messenger astronomy relies on the rapid detection of gravitational waves at very low latencies (O(1/,s)) in order to maximize the amount of time available for follow-up observations. In recent years, neural-networks have demonstrated robust non-linear modeling capabilities and millisecond-scale inference at a comparatively small computational footprint, making them an attractive family of algorithms in this context. However, integration of these algorithms into the gravitational-wave astrophysics research ecosystem has proven non-trivial. Here, we present the first fully machine learning-based pipeline for the detection of gravitational waves from compact binary coalescences (CBCs) running in low-latency. We demonstrate this pipeline to have a fraction of the latency of traditional matched filtering search pipelines while achieving state-of-the-art sensitivity to higher-mass stellar binary black holes.

The Unreasonable Ineffectiveness of the Deeper Layers
Andrey Gromov, Kushal Tirumala, Hassan Shapourian, Paolo Glorioso, Daniel A. Roberts
[ arXiv:2403.17887 ]
Foundational AI

Abstract We empirically study a simple layer-pruning strategy for popular families of open-weight pretrained LLMs, finding minimal degradation of performance on different question-answering benchmarks until after a large fraction (up to half) of the layers are removed. To prune these models, we identify the optimal block of layers to prune by considering similarity across layers; then, to "heal" the damage, we perform a small amount of finetuning. In particular, we use parameter-efficient finetuning (PEFT) methods, specifically quantization and Low Rank Adapters (QLoRA), such that each of our experiments can be performed on a single A100 GPU. From a practical perspective, these results suggest that layer pruning methods can complement other PEFT strategies to further reduce computational resources of finetuning on the one hand, and can improve the memory and latency of inference on the other hand. From a scientific perspective, the robustness of these LLMs to the deletion of layers implies either that current pretraining methods are not properly leveraging the parameters in the deeper layers of the network or that the shallow layers play a critical role in storing knowledge.

Debiasing with Diffusion: Probabilistic reconstruction of Dark Matter fields from galaxies with CAMELS
Victoria Ono, Core Francisco Park, Nayantara Mudur, Yueying Ni, Carolina Cuesta-Lazaro, Francisco Villaescusa-Navarro
The Astrophysical Journal, 2024, Volume 970, Number 2 [ arXiv:2403.10648 | code ] published
Astrophysics

Abstract Galaxies are biased tracers of the underlying cosmic web, which is dominated by dark matter components that cannot be directly observed. Galaxy formation simulations can be used to study the relationship between dark matter density fields and galaxy distributions. However, this relationship can be sensitive to assumptions in cosmology and astrophysical processes embedded in the galaxy formation models, that remain uncertain in many aspects. In this work, we develop a diffusion generative model to reconstruct dark matter fields from galaxies. The diffusion model is trained on the CAMELS simulation suite that contains thousands of state-of-the-art galaxy formation simulations with varying cosmological parameters and sub-grid astrophysics. We demonstrate that the diffusion model can predict the unbiased posterior distribution of the underlying dark matter fields from the given stellar mass fields, while being able to marginalize over uncertainties in cosmological and astrophysical models. Interestingly, the model generalizes to simulation volumes approximately 500 times larger than those it was trained on, and across different galaxy formation models. Code for reproducing these results can be found at this https URL

FeatUp: A Model-Agnostic Framework for Features at Any Resolution
Stephanie Fu, Mark Hamilton, Laura Brandt, Axel Feldman, Zhoutong Zhang, William T. Freeman
[ arXiv:2403.10516 | code ]
Foundational AI

Abstract Deep features are a cornerstone of computer vision research, capturing image semantics and enabling the community to solve downstream tasks even in the zero- or few-shot regime. However, these features often lack the spatial resolution to directly perform dense prediction tasks like segmentation and depth prediction because models aggressively pool information over large areas. In this work, we introduce FeatUp, a task- and model-agnostic framework to restore lost spatial information in deep features. We introduce two variants of FeatUp: one that guides features with high-resolution signal in a single forward pass, and one that fits an implicit model to a single image to reconstruct features at any resolution. Both approaches use a multi-view consistency loss with deep analogies to NeRFs. Our features retain their original semantics and can be swapped into existing applications to yield resolution and performance gains even without re-training. We show that FeatUp significantly outperforms other feature upsampling and image super-resolution approaches in class activation map generation, transfer learning for segmentation and depth prediction, and end-to-end training for semantic segmentation.

Supernovae Time Profiles as a Probe of New Physics at Neutrino Telescopes
Jeff Lazar, Ying-Ying Li, Carlos A. Arguelles, Vedran Brdar
[ arXiv:2403.09781 | code ]
Astrophysics

Abstract Neutrino telescopes, including IceCube, can detect galactic supernova events by observing the collective rise in photomultiplier count rates with a sub-second time resolution. Leveraging precise timing, we demonstrate the ability of neutrino telescopes to explore new weakly coupled states emitted from supernovae and subsequently decaying to neutrinos. Our approach utilizes publicly available packages, \texttt{ASTERIA} and \texttt{SNEWPY}, for simulating detector responses and parametrizing neutrino fluxes originating from Standard Model and new physics. We present results for two beyond the Standard Model scenarios and introduce the tool developed for testing a diverse range of new physics models.

Maximizing Quadruped Velocity by Minimizing Energy
Srinath Mahankali, Chi-Chang Lee, Gabriel B. Margolis, Zhang-Wei Hong, Pulkit Agrawal
2024 IEEE International Conference on Robotics and Automation [ ] published
Foundational AI

Abstract Reinforcement Learning (RL) has been a powerful tool for training robots to acquire agile locomotion skills. To learn locomotion, it is commonly necessary to introduce additional reward-shaping terms, such as an energy minimization term, to guide an algorithm like Proximal Policy Optimization (PPO) to good performance. Prior works rely on hyper-parameter tuning on the weight of the reward shaping terms to obtain satisfactory task performance. To save the efforts of tuning these weights, we adopt the Extrinsic-Intrinsic Policy Optimization (EIPO) framework. The key idea of EIPO is to establish a constrained optimization framework for the primary objective of enhancing task performance and the secondary objective of minimizing energy consumption. It seeks a policy that minimizes the energy consumption objective within the optimal policy space for task performance. This guarantees that the learned policy excels in task performance while conserving energy, all without requiring manual weight adjustments for both objectives. Our experiments evaluate EIPO on various quadruped locomotion tasks, revealing that policies trained with EIPO consistently achieve higher task performance than PPO comparisons while maintaining comparable energy consumption levels. Furthermore, EIPO exhibits superior task performance in real-world evaluations compared to PPO.

PAPERCLIP: Associating Astronomical Observations and Natural Language with Multi-Modal Models
Siddharth Mishra-Sharma, Yiding Song, Jesse Thaler
COLM 2024 [ arXiv:2403.08851 | code ] published
Astrophysics

Abstract We present PAPERCLIP (Proposal Abstracts Provide an Effective Representation for Contrastive Language-Image Pre-training), a method which associates astronomical observations imaged by telescopes with natural language using a neural network model. The model is fine-tuned from a pre-trained Contrastive Language-Image Pre-training (CLIP) model using successful observing proposal abstracts and corresponding downstream observations, with the abstracts optionally summarized via guided generation using large language models (LLMs). Using observations from the Hubble Space Telescope (HST) as an example, we show that the fine-tuned model embodies a meaningful joint representation between observations and natural language through tests targeting image retrieval (i.e., finding the most relevant observations using natural language queries) and description retrieval (i.e., querying for astrophysical object classes and use cases most relevant to a given observation). Our study demonstrates the potential for using generalist foundation models rather than task-specific models for interacting with astronomical data by leveraging text as an interface.

Moments of Clarity: Streamlining Latent Spaces in Machine Learning using Moment Pooling
Rikab Gambhir, Athis Osathapan, Jesse Thaler
Physical Review D, Volume 110, Issue 7, 1 October 2024 [ arXiv:2403.08854 | code ] published
Theoretical Physics

Abstract Many machine learning applications involve learning a latent representation of data, which is often high-dimensional and difficult to directly interpret. In this work, we propose "Moment Pooling", a natural extension of Deep Sets networks which drastically decrease latent space dimensionality of these networks while maintaining or even improving performance. Moment Pooling generalizes the summation in Deep Sets to arbitrary multivariate moments, which enables the model to achieve a much higher effective latent dimensionality for a fixed latent dimension. We demonstrate Moment Pooling on the collider physics task of quark/gluon jet classification by extending Energy Flow Networks (EFNs) to Moment EFNs. We find that Moment EFNs with latent dimensions as small as 1 perform similarly to ordinary EFNs with higher latent dimension. This small latent dimension allows for the internal representation to be directly visualized and interpreted, which in turn enables the learned internal jet representation to be extracted in closed form.

Superphot+: Realtime Fitting and Classification of Supernova Light Curves
Kaylee M. de Soto (1), Ashley Villar (1), Edo Berger (1 and 2), Sebastian Gomez (3), Griffin Hosseinzadeh (4), Doug Branton (5), Sandro Campos (6), Melissa DeLucchi (6), Jeremy Kubica (6), Olivia Lynn (6), Konstantin Malanchev (6), Alex I. Malz (6) ((1) Center for Astrophysics | Harvard & Smithsonian, (2) The NSF AI Institute for Artificial Intelligence and Fundamental Interactions, (3) Space Telescope Science Institute, (4) Steward Observatory | University of Arizona, (5) DiRAC Institute and the Department of Astronomy | University of Washington, (6) McWilliams Center for Cosmology | Department of Physics at Carnegie Mellon University)
The Astrophysics Journal, Volume 974, Number 2, 2024 [ arXiv:2403.07975 | code ] published
Astrophysics

Abstract Photometric classifications of supernova (SN) light curves have become necessary to utilize the full potential of large samples of observations obtained from wide-field photometric surveys, such as the Zwicky Transient Facility (ZTF) and the Vera C. Rubin Observatory. Here, we present a photometric classifier for SN light curves that does not rely on redshift information and still maintains comparable accuracy to redshift-dependent classifiers. Our new package, Superphot+, uses a parametric model to extract meaningful features from multiband SN light curves. We train a gradient-boosted machine with fit parameters from 6,061 ZTF SNe that pass data quality cuts and are spectroscopically classified as one of five classes: SN Ia, SN II, SN Ib/c, SN IIn, and SLSN-I. Without redshift information, our classifier yields a class-averaged F1-score of 0.61 +/- 0.02 and a total accuracy of 0.83 +/- 0.01. Including redshift information improves these metrics to 0.71 +/- 0.02 and 0.88 +/- 0.01, respectively. We assign new class probabilities to 3,558 ZTF transients that show SN-like characteristics (based on the ALeRCE Broker light curve and stamp classifiers), but lack spectroscopic classifications. Finally, we compare our predicted SN labels with those generated by the ALeRCE light curve classifier, finding that the two classifiers agree on photometric labels for 82 +/- 2% of light curves with spectroscopic labels and 72% of light curves without spectroscopic labels. Superphot+ is currently classifying ZTF SNe in real time via the ANTARES Broker, and is designed for simple adaptation to six-band Rubin light curves in the future.

Resimulation-based self-supervised learning for pretraining physics foundation models
Philip Harris, Michael Kagan, Jeffrey Krupa, Benedikt Maier, Nathaniel Woodward
Physical Review D, 2025, Volume 111, Issue 3 [ arXiv:2403.07066 ] published
Experimental Physics

Abstract Self-Supervised Learning (SSL) is at the core of training modern large machine learning models, providing a scheme for learning powerful representations that can be used in a variety of downstream tasks. However, SSL strategies must be adapted to the type of training data and downstream tasks required. We propose RS3L ('Re-simulation-based self-supervised representation learning'), a novel simulation-based SSL strategy that employs a method of re-simulation to drive data augmentation for contrastive learning in the physical sciences, particularly, in fields that rely on stochastic simulators. By intervening in the middle of the simulation process and re-running simulation components downstream of the intervention, we generate multiple realizations of an event, thus producing a set of augmentations covering all physics-driven variations available in the simulator. Using experiments from high-energy physics, we explore how this strategy may enable the development of a foundation model; we show how RS3L pre-training enables powerful performance in downstream tasks such as discrimination of a variety of objects and uncertainty mitigation. In addition to our results, we make the RS3L dataset publicly available for further studies on how to improve SSL strategies.

On classical de Sitter solutions and parametric control
David Andriot, Fabian Ruehle
Journal of High Energy Physics, Volume 2024, Article Number 101 [ arXiv:2403.07065 | code ] published
Theoretical Physics

Abstract Finding string backgrounds with de Sitter spacetime, where all approximations and corrections are controlled, is an open problem. We revisit the search for de Sitter solutions in the classical regime for specific type IIB supergravity compactifications on group manifolds, an under-explored corner of the landscape that offers an interesting testing ground for swampland conjectures. While the supergravity de Sitter solutions we obtain numerically are ambiguous in terms of their classicality, we find an analytic scaling that makes four out of six compactification radii, as well as the overall volume, arbitrarily large. This potentially provides parametric control over corrections. If we could show that these solutions, or others to be found, are fully classical, they would constitute a counterexample to conjectures stating that asymptotic de Sitter solutions do not exist. We discuss this point in great detail.

Photonic probabilistic machine learning using quantum vacuum noise
Seou Choi, Yannick Salamin, Charles Roques-Carmes, Rumen Dangovski, Di Luo, Zhuo Chen, Michael Horodynski, Jamison Sloan, Shiekh Zia Uddin, Marin Soljacic
Nature Communications, 2024, Volume 15, Article number 7760 [ arXiv:2403.04731 ] published
Theoretical Physics

Abstract Probabilistic machine learning utilizes controllable sources of randomness to encode uncertainty and enable statistical modeling. Harnessing the pure randomness of quantum vacuum noise, which stems from fluctuating electromagnetic fields, has shown promise for high speed and energy-efficient stochastic photonic elements. Nevertheless, photonic computing hardware which can control these stochastic elements to program probabilistic machine learning algorithms has been limited. Here, we implement a photonic probabilistic computer consisting of a controllable stochastic photonic element - a photonic probabilistic neuron (PPN). Our PPN is implemented in a bistable optical parametric oscillator (OPO) with vacuum-level injected bias fields. We then program a measurement-and-feedback loop for time-multiplexed PPNs with electronic processors (FPGA or GPU) to solve certain probabilistic machine learning tasks. We showcase probabilistic inference and image generation of MNIST-handwritten digits, which are representative examples of discriminative and generative models. In both implementations, quantum vacuum noise is used as a random seed to encode classification uncertainty or probabilistic generation of samples. In addition, we propose a path towards an all-optical probabilistic computing platform, with an estimated sampling rate of ~ 1 Gbps and energy consumption of ~ 5 fJ/MAC. Our work paves the way for scalable, ultrafast, and energy-efficient probabilistic machine learning hardware.

Quantum Many-Body Physics Calculations with Large Language Models
Haining Pan, Nayantara Mudur, Will Taranto, Maria Tikhanovskaya, Subhashini Venugopalan, Yasaman Bahri, Michael P. Brenner, Eun-Ah Kim
Communications Physics, 2025, Volume 8, Article number 49 [ arXiv:2403.03154 ] published
Foundational AI

Abstract Large language models (LLMs) have demonstrated an unprecedented ability to perform complex tasks in multiple domains, including mathematical and scientific reasoning. We demonstrate that with carefully designed prompts, LLMs can accurately carry out key calculations in research papers in theoretical physics. We focus on a broadly used approximation method in quantum physics: the Hartree-Fock method, requiring an analytic multi-step calculation deriving approximate Hamiltonian and corresponding self-consistency equations. To carry out the calculations using LLMs, we design multi-step prompt templates that break down the analytic calculation into standardized steps with placeholders for problem-specific information. We evaluate GPT-4's performance in executing the calculation for 15 research papers from the past decade, demonstrating that, with correction of intermediate steps, it can correctly derive the final Hartree-Fock Hamiltonian in 13 cases and makes minor errors in 2 cases. Aggregating across all research papers, we find an average score of 87.5 (out of 100) on the execution of individual calculation steps. Overall, the requisite skill for doing these calculations is at the graduate level in quantum condensed matter theory. We further use LLMs to mitigate the two primary bottlenecks in this evaluation process: (i) extracting information from papers to fill in templates and (ii) automatic scoring of the calculation steps, demonstrating good results in both cases. The strong performance is the first step for developing algorithms that automatically explore theoretical hypotheses at an unprecedented scale.

Operator Learning Renormalization Group
Xiu-Zhe Luo, Di Luo, Roger G. Melko
[ arXiv:2403.03199 | code ]
Theoretical Physics

Abstract n this paper, we present a general framework for quantum many-body simulations called the operator learning renormalization group (OLRG). Inspired by machine learning perspectives, OLRG is a generalization of Wilsons numerical renormalization group and Whites density matrix renormalization group, which recursively builds a simulatable system to approximate a target system of the same number of sites via operator maps. OLRG uses a loss function to minimize the error of a target property directly by learning the operator map in lieu of a state ansatz. This loss function is designed by a scaling consistency condition that also provides a provable bound for real-time evolution. We implement two versions of the operator maps for classical and quantum simulations. The former, which we call the Operator Matrix Map, can be implemented via neural networks on classical computers. The latter, which we call the Hamiltonian Expression Map, generates device pulse sequences to leverage the capabilities of quantum computing hardware. We illustrate the performance of both maps for calculating time-dependent quantities in the quantum Ising model Hamiltonian.

New Pathways in Neutrino Physics via Quantum-Encoded Data Analysis
Jeffrey Lazar, Santiago Giner Olavarrieta, Giancarlo Gatti, Carlos A. Argüelles, Mikel Sanz
[ arXiv:2402.19306 ]
Experimental Physics

Abstract Ever-increasing amount of data is produced by particle detectors in their quest to unveil the laws of Nature. The large data rate requires the use of specialized triggers that promptly reduce the data rate to a manageable level; however, in doing so, unexpected new phenomena may escape detection. Additionally, the large data rate is increasingly difficult to analyze effectively, which has led to a recent revolution on machine learning techniques. Here, we present a methodology based on recent quantum compression techniques that has the capacity to store exponentially more amount of information than classically available methods. To demonstrate this, we encode the full neutrino telescope event information using parity observables in an IBM quantum processor using 8 qubits. Then we show that we can recover the information stored on the quantum computer with a fidelity of 84%. Finally, we illustrate the use of our protocol by performing a classification task that separates electron-neutrino events to muon-neutrinos events in a neutrino telescope. This new capability would eventually allow us to solve the street light effect in particle physics, where we only record signatures of particles with which we are familiar.

Seeing Double: Calibrating Two Jets at Once
Rikab Gambhir, Benjamin Nachman
Physical Review D, Volume 110, Issue 7, 2024 [ arXiv:2402.14067 | code ] published
Experimental Physics

Abstract Jet energy calibration is an important aspect of many measurements and searches at the LHC. Currently, these calibrations are performed on a per-jet basis, i.e. agnostic to the properties of other jets in the same event. In this work, we propose taking advantage of the correlations induced by momentum conservation between jets in order to improve their jet energy calibration. By fitting the pT asymmetry of dijet events in simulation, while remaining agnostic to the pT spectra themselves, we are able to obtain correlation-improved maximum likelihood estimates. This approach is demonstrated with simulated jets from the CMS Detector, yielding a 3-5% relative improvement in the jet energy resolution, corresponding to a quadrature improvement of approximately 35\%.

Full-shape analysis with simulation-based priors: constraints on single field inflation from BOSS
Mikhail M. Ivanov, Carolina Cuesta-Lazaro, Siddharth Mishra-Sharma, Andrej Obuljen, Michael W. Toomey
Physical Review D, 2024, Volume 110, Issue 6 [ arXiv:2402.13310 | code ] published
Astrophysics

Abstract We present an efficient approach to set informative physically motivated priors for EFT-based full-shape analyses of galaxy survey data. We extract these priors from simulated galaxy catalogs based on halo occupation distribution (HOD) models. As a first step, we build a joint distribution between EFT galaxy bias and HOD parameters from a set of 10,500 HOD mock catalogs. We make use of the field level EFT technique that allows for cosmic variance cancellation, enabling a precision calibration of EFT parameters from computationally inexpensive small-volume simulations. As a second step, we use neural density estimators -- normalizing flows -- to model the marginal probability density of the EFT parameters, which can be used as a prior distribution in full shape analyses. As a first application, we use our HOD-based prior distribution in a new analysis of galaxy power spectra and bispectra from the BOSS survey in the context of single field primordial non-Gaussianity. We find that our approach leads to a reduction of the posterior volume of bias parameters by an order of magnitude. We also find fequilNL=650±310 and forthoNL=42±130 (at 68\% CL) in a combined two-template analysis, representing a ≈40% improvement in constraints on single field primordial non-Gaussianity, equivalent to doubling the survey volume.

Rigor with Machine Learning from Field Theory to the Poincaré Conjecture
Sergei Gukov, James Halverson, Fabian Ruehle
Nature Reviews Physics, 6, 310-319, 2024 [ arXiv:2402.13321 ] published
Theoretical Physics

Abstract Machine learning techniques are increasingly powerful, leading to many breakthroughs in the natural sciences, but they are often stochastic, error-prone, and blackbox. How, then, should they be utilized in fields such as theoretical physics and pure mathematics that place a premium on rigor and understanding? In this Perspective we discuss techniques for obtaining rigor in the natural sciences with machine learning. Non-rigorous methods may lead to rigorous results via conjecture generation or verification by reinforcement learning. We survey applications of these techniques-for-rigor ranging from string theory to the smooth 4d Poincaré conjecture in low-dimensional topology. One can also imagine building direct bridges between machine learning theory and either mathematics or theoretical physics. As examples, we describe a new approach to field theory motivated by neural network theory, and a theory of Riemannian metric flows induced by neural network gradient descent, which encompasses Perelmans formulation of the Ricci flow that was utilized to resolve the 3d Poincaré conjecture.

Real-time Dynamics of the Schwinger Model as an Open Quantum System with Neural Density Operators
Joshua Lin, Di Luo, Xiaojun Yao, Phiala E. Shanahan
Journal of High Energy Physics, Volume 2024, Article Number 211 [ arXiv:2402.06607 ] published
Theoretical Physics

Abstract Ab-initio simulations of multiple heavy quarks propagating in a Quark-Gluon Plasma are computationally difficult to perform due to the large dimension of the space of density matrices. This work develops machine learning algorithms to overcome this difficulty by approximating exact quantum states with neural network parametrisations, specifically Neural Density Operators. As a proof of principle demonstration in a QCD-like theory, the approach is applied to solve the Lindblad master equation in the 1+1d lattice Schwinger Model as an open quantum system. Neural Density Operators enable the study of in-medium dynamics on large lattice volumes, where multiple-string interactions and their effects on string-breaking and recombination phenomena can be studied. Thermal properties of the system at equilibrium can also be probed with these methods by variationally constructing the steady state of the Lindblad master equation. Scaling of this approach with system size is studied, and numerical demonstrations on up to 32 spatial lattice sites and with up to 3 interacting strings are performed.

GenEFT: Understanding statics and dynamics of model generalization via physics-inspired effective theory
David D. Baek, Ziming Liu, Max Tegmark
Physical Review E, 2025, Vol. 111, Iss. 3 [ arXiv:2402.05916 ] published
Foundational AI

Abstract We present GenEFT: an effective theory framework for shedding light on the statics and dynamics of neural network generalization, and illustrate it with graph learning examples. We first investigate the generalization phase transition as data size increases, comparing experimental results with information-theory-based approximations. We find generalization in a Goldilocks zone where the decoder is neither too weak nor too powerful. We then introduce an effective theory for the dynamics of representation learning, where latent-space representations are modeled as interacting particles (repons), and find that it explains our experimentally observed phase transition between generalization and overfitting as encoder and decoder learning rates are scanned. This highlights the power of physics-inspired effective theories for bridging the gap between theoretical predictions and practice in machine learning.

A Resource Model For Neural Scaling Law
Jinyeop Song, Ziming Liu, Max Tegmark, Jeff Gore
[ arXiv:2402.05164 ]
Foundational AI

Abstract Neural scaling laws characterize how model performance improves as the model size scales up. Inspired by empirical observations, we introduce a resource model of neural scaling. A task is usually composite hence can be decomposed into many subtasks, which compete for resources (measured by the number of neurons allocated to subtasks). On toy problems, we empirically find that: (1) The loss of a subtask is inversely proportional to its allocated neurons. (2) When multiple subtasks are present in a composite task, the resources acquired by each subtask uniformly grow as models get larger, keeping the ratios of acquired resources constants. We hypothesize these findings to be generally true and build a model to predict neural scaling laws for general composite tasks, which successfully replicates the neural scaling law of Chinchilla models reported in arXiv:2203.15556. We believe that the notion of resource used in this paper will be a useful tool for characterizing and diagnosing neural networks.

Opening the AI black box: program synthesis via mechanistic interpretability
Eric J. Michaud, Isaac Liao, Vedang Lad, Ziming Liu, Anish Mudide, Chloe Loughridge, Zifan Carl Guo, Tara Rezaei Kheirkhah, Mateja Vukelić, Max Tegmark
Entropy, Volume 26, Issue 12, 10.3390/e26121046 [ arXiv:2402.05110 | code ] published
Foundational AI

Abstract We present MIPS, a novel method for program synthesis based on automated mechanistic interpretability of neural networks trained to perform the desired task, auto-distilling the learned algorithm into Python code. We test MIPS on a benchmark of 62 algorithmic tasks that can be learned by an RNN and find it highly complementary to GPT-4: MIPS solves 32 of them, including 13 that are not solved by GPT-4 (which also solves 30). MIPS uses an integer autoencoder to convert the RNN into a finite state machine, then applies Boolean or integer symbolic regression to capture the learned algorithm. As opposed to large language models, this program synthesis technique makes no use of (and is therefore not limited by) human training data such as algorithms and code from GitHub. We discuss opportunities and challenges for scaling up this approach to make machine-learned models more interpretable and trustworthy.

LtU-ILI: An All-in-One Framework for Implicit Inference in Astrophysics and Cosmology
Matthew Ho, Deaglan J. Bartlett, Nicolas Chartier, Carolina Cuesta-Lazaro, Simon Ding, Axel Lapel, Pablo Lemos, Christopher C. Lovell, T. Lucas Makinen, Chirag Modi, Viraj Pandya, Shivam Pandey, Lucia A. Perez, Benjamin Wandelt, Greg L. Bryan
The Open Journal of Astrophysics, 2024, Volume 7 [ arXiv:2402.05137 | code ] published
Astrophysics

Abstract This paper presents the Learning the Universe Implicit Likelihood Inference (LtU-ILI) pipeline, a codebase for rapid, user-friendly, and cutting-edge machine learning (ML) inference in astrophysics and cosmology. The pipeline includes software for implementing various neural architectures, training schema, priors, and density estimators in a manner easily adaptable to any research workflow. It includes comprehensive validation metrics to assess posterior estimate coverage, enhancing the reliability of inferred results. Additionally, the pipeline is easily parallelizable, designed for efficient exploration of modeling hyperparameters. To demonstrate its capabilities, we present real applications across a range of astrophysics and cosmology problems, such as: estimating galaxy cluster masses from X-ray photometry; inferring cosmology from matter power spectra and halo point clouds; characterising progenitors in gravitational wave signals; capturing physical dust parameters from galaxy colors and luminosities; and establishing properties of semi-analytic models of galaxy formation. We also include exhaustive benchmarking and comparisons of all implemented methods as well as discussions about the challenges and pitfalls of ML inference in astronomical sciences. All code and examples are made publicly available at this https URL: https://github.com/maho3/ltu-ili.

Equivariant Symmetry Breaking Sets
YuQing Xie, Tess Smidt
[ arXiv:2402.02681 ]
Astrophysics

Abstract Equivariant neural networks (ENNs) have been shown to be extremely effective in applications involving underlying symmetries. By construction ENNs cannot produce lower symmetry outputs given a higher symmetry input. However, spontaneous symmetry breaking occurs in many physical systems and we may obtain a less symmetric stable state from an initial highly symmetric one. Hence, it is imperative that we understand how to systematically break symmetry in ENNs. In this work, we propose a novel symmetry breaking framework that is fully equivariant. We emphasize that our approach is general and applicable to equivariance under any group. To achieve this, we introduce the idea of symmetry breaking sets (SBS). Rather than redesign existing networks, we design sets of symmetry breaking objects which we feed into our network based on the symmetry of our inputs and outputs. We show there is a natural way to define equivariance on these sets, which gives an additional constraint. Minimizing the size of these sets equates to data efficiency. We prove that minimizing these sets translates to a well studied group theory problem, and tabulate solutions to this problem for the point groups. Finally, we provide some examples of symmetry breaking to demonstrate how our approach works in practice.

Substructure Detection in Realistic Strong Lensing Systems with Machine LearningSubstructure Detection in Realistic Strong Lensing Systems with Machine Learning
Arthur Tsang, Atınç Çağan Şengül, Cora Dvorkin
[ arXiv:2401.16624 ]
Astrophysics

Abstract Tens of thousands of galaxy-galaxy strong lensing systems are expected to be discovered by the end of the decade. These will form a vast new dataset that can be used to probe subgalactic dark matter structures through its gravitational effects, which will in turn allow us to study the nature of dark matter at small length scales. This work shows how we can leverage machine learning to search through the data and identify which systems are most likely to contain dark matter substructure and thus can be studied in greater depth. We use a UNet, an image segmentation architecture, on a simulated strongly-lensed dataset with realistic sources (COSMOS galaxies), lenses (power-law elliptical profiles with multipoles and external shear), and noise. Our machine learning algorithm is able to quickly detect most substructure at high image resolution and subhalo concentration. At a false positive rate of 10%, we are able to identify systems with substructure at a true positive rate of 71% for a subhalo mass range of 109-109.5M⊙. While recent detections are consistent with higher concentrations, we find that our algorithm fails at detecting subhalos with lower concentrations (expected from ΛCDM simulations).

Applications of flow models to the generation of correlated lattice QCD ensembles
Ryan Abbott, Aleksandar Botev, Denis Boyda, Daniel C. Hackett, Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2401.10874 ]
Theoretical Physics

Abstract Machine-learned normalizing flows can be used in the context of lattice quantum field theory to generate statistically correlated ensembles of lattice gauge fields at different action parameters. This work demonstrates how these correlations can be exploited for variance reduction in the computation of observables. Three different proof-of-concept applications are demonstrated using a novel residual flow architecture: continuum limits of gauge theories, the mass dependence of QCD observables, and hadronic matrix elements based on the Feynman-Hellmann approach. In all three cases, it is shown that statistical uncertainties are significantly reduced when machine-learned flows are incorporated as compared with the same calculations performed with uncorrelated ensembles or direct reweighting.

SymbolNet: neural symbolic regression with adaptive dynamic pruning for compression
Ho Fung Tsoi, Vladimir Loncar, Sridhara Dasu, Philip Harris
Machine Learning: Science and Technology, 2025, Volume 6, Number 1 [ arXiv:2401.09949 | code ] published
Foundational AI

Abstract Compact symbolic expressions have been shown to be more efficient than neural network models in terms of resource consumption and inference speed when implemented on custom hardware such as FPGAs, while maintaining comparable accuracy~cite{tsoi2023symbolic}. These capabilities are highly valuable in environments with stringent computational resource constraints, such as high-energy physics experiments at the CERN Large Hadron Collider. However, finding compact expressions for high-dimensional datasets remains challenging due to the inherent limitations of genetic programming, the search algorithm of most symbolic regression methods. Contrary to genetic programming, the neural network approach to symbolic regression offers scalability to high-dimensional inputs and leverages gradient methods for faster equation searching. Common ways of constraining expression complexity often involve multistage pruning with fine-tuning, which can result in significant performance loss. In this work, we propose SymbolNet, a neural network approach to symbolic regression specifically designed as a model compression technique, aimed at enabling low-latency inference for high-dimensional inputs on custom hardware such as FPGAs. This framework allows dynamic pruning of model weights, input features, and mathematical operators in a single training process, where both training loss and expression complexity are optimized simultaneously. We introduce a sparsity regularization term for each pruning type, which can adaptively adjust its strength, leading to convergence at a target sparsity ratio. Unlike most existing symbolic regression methods that struggle with datasets containing more than O(10) inputs, we demonstrate the effectiveness of our model on the LHC jet tagging task (16 inputs), MNIST (784 inputs), and SVHN (3072 inputs).

A Physics-Informed Variational Autoencoder for Rapid Galaxy Inference and Anomaly Detection
Alexander Gagliano, V. Ashley Villar
[ arXiv:2312.16687 | code ]
Astrophysics

Abstract The Vera C. Rubin Observatory is slated to observe nearly 20 billion galaxies during its decade-long Legacy Survey of Space and Time. The rich imaging data it collects will be an invaluable resource for probing galaxy evolution across cosmic time, characterizing the host galaxies of transient phenomena, and identifying novel populations of anomalous systems. To facilitate these studies, we introduce a convolutional variational autoencoder trained to estimate the redshift, stellar mass, and star-formation rates of galaxies from multi-band imaging data. We train and test our physics-informed CVAE on a spectroscopic sample of ∼26,000 galaxies within z<1 imaged through the Dark Energy Camera Legacy Survey. We show that our model can infer redshift and stellar mass more accurately than the latest image-based self-supervised learning approaches, and is >100x faster than more computationally-intensive SED-fitting techniques. Using a small sample of Green Pea and Red Spiral galaxies reported in the literature, we further demonstrate how this CVAE can be used to rapidly identify rare galaxy populations and interpret what makes them unique.

Applications of Lipschitz neural networks to the Run 3 LHCb trigger system
Blaise Delaney, Nicole Schulte, Gregory Ciezarek, Niklas Nolte, Mike Williams, Johannes Albrecht
[ arXiv:2312.14265 ]
Experimental Physics

Abstract The operating conditions defining the current data taking campaign at the Large Hadron Collider, known as Run 3, present unparalleled challenges for the real-time data acquisition workflow of the LHCb experiment at CERN. To address the anticipated surge in luminosity and consequent event rate, the LHCb experiment is transitioning to a fully software-based trigger system. This evolution necessitated innovations in hardware configurations, software paradigms, and algorithmic design. A significant advancement is the integration of monotonic Lipschitz neural networks into the LHCb trigger system. These deep learning models offer certified robustness against detector instabilities, and the ability to encode domain-specific inductive biases. Such properties are crucial for the inclusive heavy-flavour triggers and, most notably, for the topological triggers designed to inclusively select b-hadron candidates by exploiting the unique kinematic and decay topologies of beauty decays. This paper describes the recent progress in integrating Lipschitz neural networks into the topological triggers, highlighting the resulting enhanced sensitivity to highly displaced multi-body candidates produced within the LHCb acceptance.

First search for dark-trident processes using the MicroBooNE detector
MicroBooNE collaboration
[ arXiv:2312.13945 ]
Experimental Physics

Abstract We present a first search for dark-trident scattering in a neutrino beam using a data set corresponding to 7.2×1020 protons on target taken with the MicroBooNE detector at Fermilab. Proton interactions in the neutrino target at the Main Injector produce π0 and η mesons, which could decay into dark-matter (DM) particles mediated via a dark photon A′. A convolutional neural network is trained to identify interactions of the DM particles in the liquid-argon time projection chamber (LArTPC) exploiting its image-like reconstruction capability. In the absence of a DM signal, we provide limits at the 90% confidence level on the squared kinematic mixing parameter ε2 as a function of the dark-photon mass in the range 10≤MA′≤400 MeV. The limits cover previously unconstrained parameter space for the production of fermion or scalar DM particles χ for two benchmark models with mass ratios Mχ/MA′=0.6 and 2 and for dark fine-structure constants 0.1≤αD≤1.

Inhomogeneous energy injection in the 21-cm power spectrum: Sensitivity to dark matter decay
Yitian Sun, Joshua W. Foster, Hongwan Liu, Julian B. Muñoz, Tracy R. Slatyer
Physical Review D, Vol. 111 Iss. 4, 2025 [ arXiv:2312.11608 | code ] published
Astrophysics

Abstract The 21-cm signal provides a novel avenue to measure the thermal state of the universe during cosmic dawn and reionization (redshifts z∼5−30), and thus to probe energy injection from decaying or annihilating dark matter (DM). These DM processes are inherently inhomogeneous: both decay and annihilation are density dependent, and furthermore the fraction of injected energy that is deposited at each point depends on the gas ionization and density, leading to further anisotropies in absorption and propagation. In this work, we develop a new framework for modeling the impact of spatially inhomogeneous energy injection and deposition during cosmic dawn, accounting for ionization and baryon density dependence, as well as the attenuation of propagating photons. We showcase how this first completely inhomogeneous treatment affects the predicted 21-cm power spectrum in the presence of exotic sources of energy injection, and forecast the constraints that upcoming HERA measurements of the 21-cm power spectrum will set on DM decays to photons and to electron/positron pairs. These projected constraints considerably surpass those derived from CMB and Lyman-α measurements, and for decays to electron/positron pairs they exceed all existing constraints in the sub-GeV mass range, reaching lifetimes of ∼1028s. Our analysis demonstrates the unprecedented sensitivity of 21-cm cosmology to exotic sources of energy injection during the cosmic dark ages. Our code, DM21cm, includes all these effects and is publicly available in an accompanying release.

Cosmological Field Emulation and Parameter Inference with Diffusion Models
Nayantara Mudur, Carolina Cuesta-Lazaro, Douglas P. Finkbeiner
[ arXiv:2312.07534 ]
Astrophysics

Abstract Cosmological simulations play a crucial role in elucidating the effect of physical parameters on the statistics of fields and on constraining parameters given information on density fields. We leverage diffusion generative models to address two tasks of importance to cosmology -- as an emulator for cold dark matter density fields conditional on input cosmological parameters Ωm and σ8, and as a parameter inference model that can return constraints on the cosmological parameters of an input field. We show that the model is able to generate fields with power spectra that are consistent with those of the simulated target distribution, and capture the subtle effect of each parameter on modulations in the power spectrum. We additionally explore their utility as parameter inference models and find that we can obtain tight constraints on cosmological parameters.

Generating Interpretable Networks using Hypernetworks
Isaac Liao, Ziming Liu, Max Tegmark
[ arXiv:2312.03051 ]
Foundational AI

Abstract An essential goal in mechanistic interpretability to decode a network, i.e., to convert a neural network's raw weights to an interpretable algorithm. Given the difficulty of the decoding problem, progress has been made to understand the easier encoding problem, i.e., to convert an interpretable algorithm into network weights. Previous works focus on encoding existing algorithms into networks, which are interpretable by definition. However, focusing on encoding limits the possibility of discovering new algorithms that humans have never stumbled upon, but that are nevertheless interpretable. In this work, we explore the possibility of using hypernetworks to generate interpretable networks whose underlying algorithms are not yet known. The hypernetwork is carefully designed such that it can control network complexity, leading to a diverse family of interpretable algorithms ranked by their complexity. All of them are interpretable in hindsight, although some of them are less intuitive to humans, hence providing new insights regarding how to 'think' like a neural network. For the task of computing L1 norms, hypernetworks find three algorithms: (a) the double-sided algorithm, (b) the convexity algorithm, (c) the pudding algorithm, although only the first algorithm was expected by the authors before experiments. We automatically classify these algorithms and analyze how these algorithmic phases develop during training, as well as how they are affected by complexity control. Furthermore, we show that a trained hypernetwork can correctly construct models for input dimensions not seen in training, demonstrating systematic generalization.

What does cosmology teach us about non-gravitational properties of dark matter?
Tracy R. Slatyer
Nuclear Physics B, 2024, Volume 1003 [ ] published
Astrophysics

Abstract Cosmological observations provide our most robust evidence for dark matter that is (approximately) collisionless and cold, and furthermore can provide powerful tests of the non-gravitational properties of dark matter. There are exciting prospects for significant experimental/observational progress in the coming years. In particular, current experiments are targeting a first confirmed detection of primordial 21 cm radiation and a measurement of its power spectrum, which would open a new observational window on the end of the cosmic dark ages and cosmic dawn. On a longer timescale, there are proposed missions that could improve our measurements of the energy spectrum of the cosmic microwave background radiation by 3+ orders of magnitude, providing a new physical probe of the thermal history of the universe up to keV temperatures. In this contribution, I will discuss how signals from dark matter interactions with Standard Model particles, in particular through annihilation and decay of particle-like dark matter, could appear in these observables, and recent improvements in their theoretical modeling. There are existing stringent and broadly applicable limits on annihilating and decaying dark matter (especially at sub-GeV mass scales) from the cosmic microwave background, and complementary and competitive bounds from the Lyman-α forest for leptonically decaying light dark matter. I will outline how energy injections that are not currently excluded could change the conditions of the early universe, impact the formation of the first stars and black hole seeds, and imprint signals in the cosmological background radiation.

One-step Diffusion with Distribution Matching Distillation
Tianwei Yin, Michaël Gharbi, Richard Zhang, Eli Shechtman, Fredo Durand, William T. Freeman, Taesung Park
IEEE, 2024 [ arXiv:2311.18828 | code ] published
Foundational AI

Abstract Diffusion models generate high-quality images but require dozens of forward passes. We introduce Distribution Matching Distillation (DMD), a procedure to transform a diffusion model into a one-step image generator with minimal impact on image quality. We enforce the one-step image generator match the diffusion model at distribution level, by minimizing an approximate KL divergence whose gradient can be expressed as the difference between 2 score functions, one of the target distribution and the other of the synthetic distribution being produced by our one-step generator. The score functions are parameterized as two diffusion models trained separately on each distribution. Combined with a simple regression loss matching the large-scale structure of the multi-step diffusion outputs, our method outperforms all published few-step diffusion approaches, reaching 2.62 FID on ImageNet 64x64 and 11.49 FID on zero-shot COCO-30k, comparable to Stable Diffusion but orders of magnitude faster. Utilizing FP16 inference, our model generates images at 20 FPS on modern hardware.

Anomaly Detection in Collider Physics via Factorized Observables
Eric M. Metodiev, Jesse Thaler, Raymond Wynne
Physical Review D, Volume 110, Issue 5, 1 September 2024 [ arXiv:2312.00119 | code ] published
Theoretical Physics

Abstract To maximize the discovery potential of high-energy colliders, experimental searches should be sensitive to unforeseen new physics scenarios. This goal has motivated the use of machine learning for unsupervised anomaly detection. In this paper, we introduce a new anomaly detection strategy called FORCE: factorized observables for regressing conditional expectations. Our approach is based on the inductive bias of factorization, which is the idea that the physics governing different energy scales can be treated as approximately independent. Assuming factorization holds separately for signal and background processes, the appearance of non-trivial correlations between low- and high-energy observables is a robust indicator of new physics. Under the most restrictive form of factorization, a machine-learned model trained to identify such correlations will in fact converge to the optimal new physics classifier. We test FORCE on a benchmark anomaly detection task for the Large Hadron Collider involving collimated sprays of particles called jets. By teasing out correlations between the kinematics and substructure of jets, our method can reliably extract percent-level signal fractions. This strategy for uncovering new physics adds to the growing toolbox of anomaly detection methods for collider physics with a complementary set of assumptions.

Learning an Effective Evolution Equation for Particle-Mesh Simulations Across Cosmologies
Nicolas Payot, Pablo Lemos, Laurence Perreault-Levasseur, Carolina Cuesta-Lazaro, Chirag Modi, Yashar Hezaveh
[ arXiv:2311.18017 ]
Astrophysics

Abstract Particle-mesh simulations trade small-scale accuracy for speed compared to traditional, computationally expensive N-body codes in cosmological simulations. In this work, we show how a data-driven model could be used to learn an effective evolution equation for the particles, by correcting the errors of the particle-mesh potential incurred on small scales during simulations. We find that our learnt correction yields evolution equations that generalize well to new, unseen initial conditions and cosmologies. We further demonstrate that the resulting corrected maps can be used in a simulation-based inference framework to yield an unbiased inference of cosmological parameters. The model, a network implemented in Fourier space, is exclusively trained on the particle positions and velocities.

Growing Brains in Recurrent Neural Networks for Multiple Cognitive Tasks
Ziming Liu, Mikail Khona, Ila Fiete, Max Tegmark
NeurIPS 2023 Workshop NeurReps [ ] published
Foundational AI

Abstract Recurrent neural networks (RNNs) trained on a diverse ensemble of cognitive tasks, as described by Yang et al (2019); Khona et al. (2023), have been shown to exhibit functional modularity, where neurons organize into discrete functional clusters, each specialized for specific shared computational subtasks. However, these RNNs do not demonstrate anatomical modularity, where these functionally specialized clusters also have a distinct spatial organization. This contrasts with the human brain which has both functional and anatomical modularity. Is there a way to train RNNs to make them more like brains in this regard? We apply a recent machine learning method, brain-inspired modular training (BIMT), to encourage neural connectivity to be local in space. Consequently, hidden neuron organization of the RNN forms spatial structures reminiscent of those of the brain: spatial clusters which correspond to functional clusters. Compared to standard regularization and absence of regularization, BIMT exhibits superior performance by optimally balancing between task performance and sparsity. This balance is quantified both in terms of the number of active neurons and the cumulative wiring length. In addition to achieving brain-like organization in RNNs, our findings also suggest that BIMT holds promise for applications in neuromorphic computing and enhancing the interpretability of neural network architectures.

A point cloud approach to generative modeling for galaxy surveys at the field level
Carolina Cuesta-Lazaro, Siddharth Mishra-Sharma
[ arXiv:2311.17141 | code ]
Astrophysics

Abstract We introduce a diffusion-based generative model to describe the distribution of galaxies in our Universe directly as a collection of points in 3-D space (coordinates) optionally with associated attributes (e.g., velocities and masses), without resorting to binning or voxelization. The custom diffusion model can be used both for emulation, reproducing essential summary statistics of the galaxy distribution, as well as inference, by computing the conditional likelihood of a galaxy field. We demonstrate a first application to massive dark matter haloes in the Quijote simulation suite. This approach can be extended to enable a comprehensive analysis of cosmological data, circumventing limitations inherent to summary statistic -- as well as neural simulation-based inference methods.

Symphony: Symmetry-Equivariant Point-Centered Spherical Harmonics for Molecule Generation
Ameya Daigavane, Song Kim, Mario Geiger, Tess Smidt
[ arXiv:2311.16199 | code ]
Foundational AI

Abstract We present Symphony, an E(3)-equivariant autoregressive generative model for 3D molecular geometries that iteratively builds a molecule from molecular fragments. Existing autoregressive models such as G-SchNet and G-SphereNet for molecules utilize rotationally invariant features to respect the 3D symmetries of molecules. In contrast, Symphony uses message-passing with higher-degree E(3)-equivariant features. This allows a novel representation of probability distributions via spherical harmonic signals to efficiently model the 3D geometry of molecules. We show that Symphony is able to accurately generate small molecules from the QM9 dataset, outperforming existing autoregressive models and approaching the performance of diffusion models.

Probabilistic reconstruction of Dark Matter fields from biased tracers using diffusion models
Core Francisco Park, Victoria Ono, Nayantara Mudur, Yueying Ni, Carolina Cuesta-Lazaro
[ arXiv:2311.08558 | code ]
Astrophysics

Abstract Galaxies are biased tracers of the underlying cosmic web, which is dominated by dark matter components that cannot be directly observed. The relationship between dark matter density fields and galaxy distributions can be sensitive to assumptions in cosmology and astrophysical processes embedded in the galaxy formation models, that remain uncertain in many aspects. Based on state-of-the-art galaxy formation simulation suites with varied cosmological parameters and sub-grid astrophysics, we develop a diffusion generative model to predict the unbiased posterior distribution of the underlying dark matter fields from the given stellar mass fields, while being able to marginalize over the uncertainties in cosmology and galaxy formation.

Safe but Incalculable: Energy-weighting is not all you need
Samuel Bright-Thonney, Benjamin Nachman, Jesse Thaler
Physical Review D, 2024, Volume 110, Issue 1 [ arXiv:2311.07652 ] published
Theoretical Physics

Abstract Infrared and collinear (IRC) safety has long been used a proxy for robustness when developing new jet substructure observables. This guiding philosophy has been carried into the deep learning era, where IRC-safe neural networks have been used for many jet studies. For graph-based neural networks, the most straightforward way to achieve IRC safety is to weight particle inputs by their energies. However, energy-weighting by itself does not guarantee that perturbative calculations of machine-learned observables will enjoy small non-perturbative corrections. In this paper, we demonstrate the sensitivity of IRC-safe networks to non-perturbative effects, by training an energy flow network (EFN) to maximize its sensitivity to hadronization. We then show how to construct Lipschitz Energy Flow Networks (L-EFNs), which are both IRC safe and relatively insensitive to non-perturbative corrections. We demonstrate the performance of L-EFNs on generated samples of quark and gluon jets, and showcase fascinating differences between the learned latent representations of EFNs and L-EFNs.

Two Watts is All You Need: Enabling In-Detector Real-Time Machine Learning for Neutrino Telescopes Via Edge Computing
Miaochen Jin, Yushi Hu, Carlos A. Argüelles
[ arXiv:2311.04983 ]
Experimental Physics

Abstract The use of machine learning techniques has significantly increased the physics discovery potential of neutrino telescopes. In the upcoming years, we are expecting upgrade of currently existing detectors and new telescopes with novel experimental hardware, yielding more statistics as well as more complicated data signals. This calls out for an upgrade on the software side needed to handle this more complex data in a more efficient way. Specifically, we seek low power and fast software methods to achieve real-time signal processing, where current machine learning methods are too expensive to be deployed in the resource-constrained regions where these experiments are located. We present the first attempt at and a proof-of-concept for enabling machine learning methods to be deployed in-detector for water/ice neutrino telescopes via quantization and deployment on Google Edge Tensor Processing Units (TPUs). We design a recursive neural network with a residual convolutional embedding, and adapt a quantization process to deploy the algorithm on a Google Edge TPU. This algorithm can achieve similar reconstruction accuracy compared with traditional GPU-based machine learning solutions while requiring the same amount of power compared with CPU-based regression solutions, combining the high accuracy and low power advantages and enabling real-time in-detector machine learning in even the most power-restricted environments.

T-Duality and Flavor Symmetries in Little String Theories
Hamza Ahmed, Paul-Konstantin Oehlmann, Fabian Ruehle
Journal of High Energy Physics, Volume 2024, article number 61 [ arXiv:2311.02168 ] published
Theoretical Physics

Abstract We explore the T-duality web of 6D Heterotic Little String Theories, focusing on flavor algebra reducing deformations. A careful analysis of the full flavor algebra, including Abelian factors, shows that the flavor rank is preserved under T-duality. This suggests a new T-duality invariant in addition to the Coulomb branch dimension and the two-group structure constants. We also engineer Little String Theories with non-simply laced flavor algebras, whose appearance we attribute to certain discrete 3-form fluxes in M-theory. Geometrically, these theories are engineered in F-theory with non-Kähler favorable K3 fibers. This geometric origin leads us to propose that freezing fluxes are preserved across T-duality. Along the way, we discuss various exotic models, including two inequivalent Spin(32)/ℤ2 models that are dual to the same E8×E8 theory, and a family of self-T-dual models.

Pairing-based graph neural network for simulating quantum materials
Di Luo, David D. Dai, Liang Fu
[ arXiv:2311.02143 ]
Astrophysics

Abstract We introduce a pairing-based graph neural network, GemiNet, for simulating quantum many-body systems. Our architecture augments a BCS mean-field wavefunction with a generalized pair amplitude parameterized by a graph neural network. Variational Monte Carlo with GemiNet simultaneously provides an accurate, flexible, and scalable method for simulating many-electron systems. We apply GemiNet to two-dimensional semiconductor electron-hole bilayers and obtain highly accurate results on a variety of interaction-induced phases, including the exciton Bose-Einstein condensate, electron-hole superconductor, and bilayer Wigner crystal. Our study demonstrates the potential of physically-motivated neural network wavefunctions for quantum materials simulations.

Learning to See Physical Properties with Active Sensing Motor Policies
Gabriel B. Margolis, Xiang Fu, Yandong Ji, Pulkit Agrawal
[ arXiv:2311.01405 | code ]
Foundational AI

Abstract Knowledge of terrain's physical properties inferred from color images can aid in making efficient robotic locomotion plans. However, unlike image classification, it is unintuitive for humans to label image patches with physical properties. Without labeled data, building a vision system that takes as input the observed terrain and predicts physical properties remains challenging. We present a method that overcomes this challenge by self-supervised labeling of images captured by robots during real-world traversal with physical property estimators trained in simulation. To ensure accurate labeling, we introduce Active Sensing Motor Policies (ASMP), which are trained to explore locomotion behaviors that increase the accuracy of estimating physical parameters. For instance, the quadruped robot learns to swipe its foot against the ground to estimate the friction coefficient accurately. We show that the visual system trained with a small amount of real-world traversal data accurately predicts physical parameters. The trained system is robust and works even with overhead images captured by a drone despite being trained on data collected by cameras attached to a quadruped robot walking on the ground.

E(2) Equivariant Neural Networks for Robust Galaxy Morphology Classification
Sneh Pandya, Purvik Patel, Franc O, Jonathan Blazek
[ arXiv:2311.01500 | code ]
Astrophysics

Abstract We propose the use of group convolutional neural network architectures (GCNNs) equivariant to the 2D Euclidean group, E(2), for the task of galaxy morphology classification by utilizing symmetries of the data present in galaxy images as an inductive bias in the architecture. We conduct robustness studies by introducing artificial perturbations via Poisson noise insertion and one-pixel adversarial attacks to simulate the effects of limited observational capabilities. We train, validate, and test GCNNs equivariant to discrete subgroups of E(2) - the cyclic and dihedral groups of order N - on the Galaxy10 DECals dataset and find that GCNNs achieve higher classification accuracy and are consistently more robust than their non-equivariant counterparts, with an architecture equivariant to the group D16 achieving a 95.52±0.18% test-set accuracy. We also find that the model loses <6% accuracy on a 50%-noise dataset and all GCNNs are less susceptible to one-pixel perturbations than an identically constructed CNN..

Metric Flows with Neural Networks
James Halverson, Fabian Ruehle
Machine Learning: Science and Technology, 2024, Volume 5, Number 4 [ arXiv:2310.19870 ] published
Theoretical Physics

Abstract We develop a theory of flows in the space of Riemannian metrics induced by neural network gradient descent. This is motivated in part by recent advances in approximating Calabi-Yau metrics with neural networks and is enabled by recent advances in understanding flows in the space of neural networks. We derive the corresponding metric flow equations, which are governed by a metric neural tangent kernel, a complicated, non-local object that evolves in time. However, many architectures admit an infinite-width limit in which the kernel becomes fixed and the dynamics simplify. Additional assumptions can induce locality in the flow, which allows for the realization of Perelman's formulation of Ricci flow that was used to resolve the 3d Poincaré conjecture. We apply these ideas to numerical Calabi-Yau metrics, including a discussion on the importance of feature learning.

Precise Cosmological Constraints from BOSS Galaxy Clustering with a Simulation-Based Emulator of the Wavelet Scattering Transform
Georgios Valogiannis, Sihan Yuan, Cora Dvorkin
Physical Review D 2024, Volume 109, Issue 10 [ arXiv:2310.16116 ] published
Astrophysics

Abstract We perform a reanalysis of the BOSS CMASS DR12 galaxy dataset using a simulation-based emulator for the Wavelet Scattering Transform (WST) coefficients. Moving beyond our previous works, which laid the foundation for the first galaxy clustering application of this estimator, we construct a neural net-based emulator for the cosmological dependence of the WST coefficients and the 2-point correlation function multipoles, trained from the state-of-the-art suite of extsc{AbacusSummit} simulations combined with a flexible Halo Occupation Distribution (HOD) galaxy model. In order to confirm the accuracy of our pipeline, we subject it to a series of thorough internal and external mock parameter recovery tests, before applying it to reanalyze the CMASS observations in the redshift range 0.46<z<0.57. We find that a joint WST + 2-point correlation function likelihood analysis allows us to obtain marginalized 1σ errors on the ΛCDM parameters that are tighter by a factor of 2.5−6, compared to the 2-point correlation function, and by a factor of 1.4−2.5 compared to the WST-only results. This corresponds to a competitive 0.9%, 2.3% and 1% level of determination for parameters ωc, σ8 & ns, respectively, and also to a 0.7% & 2.5% constraint on derived parameters h and f(z)σ8(z), in agreement with the extit{Planck} 2018 results. Our results reaffirm the constraining power of the WST and highlight the exciting prospect of employing higher-order statistics in order to fully exploit the power of upcoming Stage-IV spectroscopic observations.

Functional renormalization group for signal detection and stochastic ergodicity breaking
Harold Erbin, Riccardo Finotello, Bio Wahabou Kpera, Vincent Lahoche, Dine Ousmane Samary
Journal of Statistical Mechanics: Theory and Experiment, Volume 2024 [ arXiv:2310.07499 ] published
Theoretical Physics

Abstract Signal detection is one of the main challenges of data science. As it often happens in data analysis, the signal in the data may be corrupted by noise. There is a wide range of techniques aimed at extracting the relevant degrees of freedom from data. However, some problems remain difficult. It is notably the case of signal detection in almost continuous spectra when the signal-to-noise ratio is small enough. This paper follows a recent bibliographic line which tackles this issue with field-theoretical methods. Previous analysis focused on equilibrium Boltzmann distributions for some effective field representing the degrees of freedom of data. It was possible to establish a relation between signal detection and ℤ2-symmetry breaking. In this paper, we consider a stochastic field framework inspiring by the so-called 'Model A', and show that the ability to reach or not an equilibrium state is correlated with the shape of the dataset. In particular, studying the renormalization group of the model, we show that the weak ergodicity prescription is always broken for signals small enough, when the data distribution is close to the Marchenko-Pastur (MP) law. This, in particular, enables the definition of a detection threshold in the regime where the signal-to-noise ratio is small enough.

Growing Brains: Co-emergence of Anatomical and Functional Modularity in Recurrent Neural Networks
Ziming Liu, Mikail Khona, Ila R. Fiete, Max Tegmark
[ arXiv:2310.07711 ]
Foundational AI

Abstract Recurrent neural networks (RNNs) trained on compositional tasks can exhibit functional modularity, in which neurons can be clustered by activity similarity and participation in shared computational subtasks. Unlike brains, these RNNs do not exhibit anatomical modularity, in which functional clustering is correlated with strong recurrent coupling and spatial localization of functional clusters. Contrasting with functional modularity, which can be ephemerally dependent on the input, anatomically modular networks form a robust substrate for solving the same subtasks in the future. To examine whether it is possible to grow brain-like anatomical modularity, we apply a recent machine learning method, brain-inspired modular training (BIMT), to a network being trained to solve a set of compositional cognitive tasks. We find that functional and anatomical clustering emerge together, such that functionally similar neurons also become spatially localized and interconnected. Moreover, compared to standard L1 or no regularization settings, the model exhibits superior performance by optimally balancing task performance and network sparsity. In addition to achieving brain-like organization in RNNs, our findings also suggest that BIMT holds promise for applications in neuromorphic computing and enhancing the interpretability of neural network architectures.

Search for heavy neutral leptons in electron-positron and neutral-pion final states with the MicroBooNE detector
MicroBooNE collaboration
[ arXiv:2310.07660 ]
Experimental Physics

Abstract We present the first search for heavy neutral leptons (HNL) decaying into νe+e− or νπ0 final states in a liquid-argon time projection chamber using data collected with the MicroBooNE detector. The data were recorded synchronously with the NuMI neutrino beam from Fermilab's Main Injector corresponding to a total exposure of 7.01×1020 protons on target. We set upper limits at the 90% confidence level on the mixing parameter |Uμ4|2 in the mass ranges 10≤mHNL≤150 MeV for the νe+e− channel and 150≤mHNL≤245 MeV for the νπ0 channel, assuming |Ue4|2=|Uτ4|2=0. These limits represent the most stringent constraints in the mass range 35<mHNL<175 MeV and the first constraints from a direct search for νπ0 decays.

The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets
Samuel Marks, Max Tegmark
[ arXiv:2310.06824 | code ]
Foundational AI

Abstract Large Language Models (LLMs) have impressive capabilities, but are also prone to outputting falsehoods. Recent work has developed techniques for inferring whether a LLM is telling the truth by training probes on the LLM's internal activations. However, this line of work is controversial, with some authors pointing out failures of these probes to generalize in basic ways, among other conceptual issues. In this work, we curate high-quality datasets of true/false statements and use them to study in detail the structure of LLM representations of truth, drawing on three lines of evidence: 1. Visualizations of LLM true/false statement representations, which reveal clear linear structure. 2. Transfer experiments in which probes trained on one dataset generalize to different datasets. 3. Causal evidence obtained by surgically intervening in a LLM's forward pass, causing it to treat false statements as true and vice versa. Overall, we present evidence that language models linearly represent the truth or falsehood of factual statements. We also introduce a novel technique, mass-mean probing, which generalizes better and is more causally implicated in model outputs than other probing techniques.

Grokking as Compression: A Nonlinear Complexity Perspective
Ziming Liu, Ziqian Zhong, Max Tegmark
[ arXiv:2310.05918 ]
Foundational AI

Abstract We attribute grokking, the phenomenon where generalization is much delayed after memorization, to compression. To do so, we define linear mapping number (LMN) to measure network complexity, which is a generalized version of linear region number for ReLU networks. LMN can nicely characterize neural network compression before generalization. Although the L2 norm has been a popular choice for characterizing model complexity, we argue in favor of LMN for a number of reasons: (1) LMN can be naturally interpreted as information/computation, while L2 cannot. (2) In the compression phase, LMN has linear relations with test losses, while L2 is correlated with test losses in a complicated nonlinear way. (3) LMN also reveals an intriguing phenomenon of the XOR network switching between two generalization solutions, while L2 does not. Besides explaining grokking, we argue that LMN is a promising candidate as the neural network version of the Kolmogorov complexity since it explicitly considers local or conditioned linear computations aligned with the nature of modern artificial neural networks.

Discovering Symmetry Breaking in Physical Systems with Relaxed Group Convolution
Rui Wang, Elyssa Hofgard, Han Gao, Robin Walters, Tess E. Smidt
International Conference on Machine Learning, 2024 [ arXiv:2310.02299 ] published
Foundational AI

Abstract Modeling symmetry breaking is essential for understanding the fundamental changes in the behaviors and properties of physical systems, from microscopic particle interactions to macroscopic phenomena like fluid dynamics and cosmic structures. Thus, identifying sources of asymmetry is an important tool for understanding physical systems. In this paper, we focus on learning asymmetries of data using relaxed group convolutions. We provide both theoretical and empirical evidence that this flexible convolution technique allows the model to maintain the highest level of equivariance that is consistent with data and discover the subtle symmetry-breaking factors in various physical systems. We employ various relaxed group convolution architectures to uncover various symmetry-breaking factors that are interpretable and physically meaningful in different physical systems, including the phase transition of crystal structure, the isotropy and homogeneity breaking in turbulent flow, and the time-reversal symmetry breaking in pendulum systems.

A Neural Scaling Law from Lottery Ticket Ensembling
Ziming Liu, Max Tegmark
[ arXiv:2310.02258 ]
Foundational AI

Abstract Neural scaling laws (NSL) refer to the phenomenon where model performance improves with scale. Sharma & Kaplan analyzed NSL using approximation theory and predict that MSE losses decay as N−α, α=4/d, where N is the number of model parameters, and d is the intrinsic input dimension. Although their theory works well for some cases (e.g., ReLU networks), we surprisingly find that a simple 1D problem y=x2 manifests a different scaling law (α=1) from their predictions (α=4). We opened the neural networks and found that the new scaling law originates from lottery ticket ensembling: a wider network on average has more 'lottery tickets', which are ensembled to reduce the variance of outputs. We support the ensembling mechanism by mechanistically interpreting single neural networks, as well as studying them statistically. We attribute the N−1 scaling law to the 'central limit theorem' of lottery tickets. Finally, we discuss its potential implications for large language models and statistical physics-type theories of learning.

Cosmological constraints from density-split clustering in the BOSS CMASS galaxy sample
Enrique Paillas, Carolina Cuesta-Lazaro, Will J. Percival, Seshadri Nadathur, Yan-Chuan Cai, Sihan Yuan, Florian Beutler, Arnaud de Mattia, Daniel Eisenstein, Daniel Forero-Sanchez, Nelson Padilla, Mathilde Pinon, Vanina Ruhlmann-Kleider, Ariel G. Sánchez, Georgios Valogiannis, Pauline Zarrouk
[ arXiv:2309.16541 | code ]
Astrophysics

Abstract We present a clustering analysis of the BOSS DR12 CMASS galaxy sample, combining measurements of the galaxy two-point correlation function and density-split clustering down to a scale of 1h−1Mpc. Our theoretical framework is based on emulators trained on high-fidelity mock galaxy catalogues that forward model the cosmological dependence of the clustering statistics within an extended-ΛCDM framework, including redshift-space and Alcock-Paczynski distortions. Our base-ΛCDM analysis finds ωcdm=0.1201±0.0022, σ8=0.792±0.034, and ns=0.970±0.018, corresponding to fσ8=0.462±0.020 at z≈0.525, which is in agreement with Planck 2018 predictions and various clustering studies in the literature. We test single-parameter extensions to base-ΛCDM, varying the running of the spectral index, the dark energy equation of state, and the density of massless relic neutrinos, finding no compelling evidence for deviations from the base model. We model the galaxy-halo connection using a halo occupation distribution framework, finding signatures of environment-based assembly bias in the data. We validate our pipeline against mock catalogues that match the clustering and selection properties of CMASS, showing that we can recover unbiased cosmological constraints even with a volume 84 times larger than the one used in this study.

SUNBIRD: A simulation-based model for full-shape density-split clustering
Carolina Cuesta-Lazaro, Enrique Paillas, Sihan Yuan, Yan-Chuan Cai, Seshadri Nadathur, Will J. Percival, Florian Beutler, Arnaud de Mattia, Daniel Eisenstein, Daniel Forero-Sanchez, Nelson Padilla, Mathilde Pinon, Vanina Ruhlmann-Kleider, Ariel G. Sánchez, Georgios Valogiannis, Pauline Zarrouk
[ arXiv:2309.16539 | code ]
Astrophysics

Abstract Combining galaxy clustering information from regions of different environmental densities can help break cosmological parameter degeneracies and access non-Gaussian information from the density field that is not readily captured by the standard two-point correlation function (2PCF) analyses. However, modelling these density-dependent statistics down to the non-linear regime has so far remained challenging. We present a simulation-based model that is able to capture the cosmological dependence of the full shape of the density-split clustering (DSC) statistics down to intra-halo scales. Our models are based on neural-network emulators that are trained on high-fidelity mock galaxy catalogues within an extended-ΛCDM framework, incorporating the effects of redshift-space, Alcock-Paczynski distortions and models of the halo-galaxy connection. Our models reach sub-percent level accuracy down to 1h−1Mpc and are robust against different choices of galaxy-halo connection modelling. When combined with the galaxy 2PCF, DSC can tighten the constraints on ωcdm, σ8, and ns by factors of 2.9, 1.9, and 2.1, respectively, compared to a 2PCF-only analysis. DSC additionally puts strong constraints on environment-based assembly bias parameters. Our code is made publicly available on Github.

Chained Quantile Morphing with Normalizing Flows
Samuel Bright-Thonney, Philip Harris, Patrick McCormack, Simon Rothman
[ arXiv:2309.15912 ]
Experimental Physics

Abstract Accounting for inaccuracies in Monte Carlo simulations is a crucial step in any high energy physics analysis. It becomes especially important when training machine learning models, which can amplify simulation inaccuracies and introduce large discrepancies and systematic uncertainties when the model is applied to data. In this paper, we introduce a method to transform simulated events to better match data using normalizing flows, a class of deep learning-based density estimation models. Our proposal uses a technique called chained quantile morphing, which corrects a set of observables by iteratively shifting each entry according to a conditonal cumulative density function. We demonstrate the technique on a realistic particle physics dataset, and compare it to a neural network-based reweighting method. We also introduce a new contrastive learning technique to correct high dimensional particle-level inputs, which naively cannot be efficiently corrected with morphing strategies.

GWAK: Gravitational-Wave Anomalous Knowledge with Recurrent Autoencoders
Ryan Raikman, Eric A. Moreno, Ekaterina Govorkova, Ethan J Marx, Alec Gunny, William Benoit, Deep Chatterjee, Rafia Omer, Muhammed Saleem, Dylan S Rankin, Michael W Coughlin, Philip C Harris, Erik Katsavounidis
Journal of High Energy Physics 2024, Volume 2024, Article number 158 [ arXiv:2309.11537 | code ] published
Experimental Physics

Abstract Matched-filtering detection techniques for gravitational-wave (GW) signals in ground-based interferometers rely on having well-modeled templates of the GW emission. Such techniques have been traditionally used in searches for compact binary coalescences (CBCs), and have been employed in all known GW detections so far. However, interesting science cases aside from compact mergers do not yet have accurate enough modeling to make matched filtering possible, including core-collapse supernovae and sources where stochasticity may be involved. Therefore the development of techniques to identify sources of these types is of significant interest. In this paper, we present a method of anomaly detection based on deep recurrent autoencoders to enhance the search region to unmodeled transients. We use a semi-supervised strategy that we name Gravitational Wave Anomalous Knowledge (GWAK). While the semi-supervised nature of the problem comes with a cost in terms of accuracy as compared to supervised techniques, there is a qualitative advantage in generalizing experimental sensitivity beyond pre-computed signal templates. We construct a low-dimensional embedded space using the GWAK method, capturing the physical signatures of distinct signals on each axis of the space. By introducing signal priors that capture some of the salient features of GW signals, we allow for the recovery of sensitivity even when an unmodeled anomaly is encountered. We show that regions of the GWAK space can identify CBCs, detector glitches and also a variety of unmodeled astrophysical sources.

Simulation-based Inference for Exoplanet Atmospheric Retrieval: Insights from winning the Ariel Data Challenge 2023 using Normalizing Flows
Mayeul Aubin, Carolina Cuesta-Lazaro, Ethan Tregidga, Javier Viaña, Cecilia Garraffo, Iouli E. Gordon, Mercedes López-Morales, Robert J. Hargreaves, Vladimir Yu. Makhnev, Jeremy J. Drake, Douglas P. Finkbeiner, Phillip Cargile
[ arXiv:2309.09337 | code ]
Astrophysics

Abstract Advancements in space telescopes have opened new avenues for gathering vast amounts of data on exoplanet atmosphere spectra. However, accurately extracting chemical and physical properties from these spectra poses significant challenges due to the non-linear nature of the underlying physics. This paper presents novel machine learning models developed by the AstroAI team for the Ariel Data Challenge 2023, where one of the models secured the top position among 293 competitors. Leveraging Normalizing Flows, our models predict the posterior probability distribution of atmospheric parameters under different atmospheric assumptions. Moreover, we introduce an alternative model that exhibits higher performance potential than the winning model, despite scoring lower in the challenge. These findings highlight the need to reevaluate the evaluation metric and prompt further exploration of more efficient and accurate approaches for exoplanet atmosphere spectra analysis. Finally, we present recommendations to enhance the challenge and models, providing valuable insights for future applications on real observational data. These advancements pave the way for more effective and timely analysis of exoplanet atmospheric properties, advancing our understanding of these distant worlds.

Mitigating Confirmation Bias in Semi-supervised Learning via Efficient Bayesian Model Averaging
Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, Seungwook Han, Ligong Han, Leonid Karlinsky, Marin Soljacic, Akash Srivastava
Transactions on Machine Learning Research 2023, Submission number 1013 [ | code ] published
Foundational AI

Abstract State-of-the-art (SOTA) semi-supervised learning (SSL) methods have been highly successful in leveraging a mix of labeled and unlabeled data, often via self-training or pseudo-labeling. During pseudo-labeling, the model's predictions on unlabeled data are used for training and may result in confirmation bias where the model reinforces its own mistakes. In this work, we show that SOTA SSL methods often suffer from confirmation bias and demonstrate that this is often a result of using a poorly calibrated classifier for pseudo labeling. We introduce BaM-SSL, an efficient Bayesian Model averaging technique that improves uncertainty quantification in SSL methods with limited computational or memory overhead. We demonstrate that BaM-SSL mitigates confirmation bias in SOTA SSL methods across standard vision benchmarks of CIFAR-10, CIFAR-100, giving up to 16% improvement in test accuracy on the CIFAR-100 with 400 labels benchmark. Furthermore, we also demonstrate their effectiveness in additional realistic and challenging problems, such as class-imbalanced datasets and in photonics science.

Advances in machine-learning-based sampling motivated by lattice quantum chromodynamics
Kyle Cranmer, Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Phiala E. Shanahan
Nature Reviews Physics, 2023, Volume 5 [ arXiv:2309.01156 ] published
Theoretical Physics

Abstract Sampling from known probability distributions is a ubiquitous task in computational science, underlying calculations in domains from linguistics to biology and physics. Generative machine-learning (ML) models have emerged as a promising tool in this space, building on the success of this approach in applications such as image, text, and audio generation. Often, however, generative tasks in scientific domains have unique structures and features -- such as complex symmetries and the requirement of exactness guarantees -- that present both challenges and opportunities for ML. This Perspective outlines the advances in ML-based sampling motivated by lattice quantum field theory, in particular for the theory of quantum chromodynamics. Enabling calculations of the structure and interactions of matter from our most fundamental understanding of particle physics, lattice quantum chromodynamics is one of the main consumers of open-science supercomputing worldwide. The design of ML algorithms for this application faces profound challenges, including the necessity of scaling custom ML architectures to the largest supercomputers, but also promises immense benefits, and is spurring a wave of development in ML-based sampling more broadly. In lattice field theory, if this approach can realize its early promise it will be a transformative step towards first-principles physics calculations in particle, nuclear and condensed matter physics that are intractable with traditional approaches.

Signal-to-noise improvement through neural network contour deformations for 3D SU(2) lattice gauge theory
William Detmold, Gurtej Kanwar, Yin Lin, Phiala E. Shanahan, Michael L. Wagman
[ arXiv:2309.00600 ]
Theoretical Physics

Abstract Complex contour deformations of the path integral have been demonstrated to significantly improve the signal-to-noise ratio of observables in previous studies of two-dimensional gauge theories with open boundary conditions. In this work, new developments based on gauge fixing and a neural network definition of the deformation are introduced, which enable an effective application to theories in higher dimensions and with generic boundary conditions. Improvements of the signal-to-noise ratio by up to three orders of magnitude for Wilson loop measurements are shown in SU(2) lattice gauge theory in three spacetime dimensions.

Reconstructing S-matrix Phases with Machine Learning
Aurélien Dersy, Matthew D. Schwartz, Alexander Zhiboedov
Journal of High Energy Physics, Volume 2024, Article 200 [ arXiv:2308.09451 | code ] published
Theoretical Physics

Abstract An important element of the S-matrix bootstrap program is the relationship between the modulus of an S-matrix element and its phase. Unitarity relates them by an integral equation. Even in the simplest case of elastic scattering, this integral equation cannot be solved analytically and numerical approaches are required. We apply modern machine learning techniques to studying the unitarity constraint. We find that for a given modulus, when a phase exists it can generally be reconstructed to good accuracy with machine learning. Moreover, the loss of the reconstruction algorithm provides a good proxy for whether a given modulus can be consistent with unitarity at all. In addition, we study the question of whether multiple phases can be consistent with a single modulus, finding novel phase-ambiguous solutions. In particular, we find a new phase-ambiguous solution which pushes the known limit on such solutions significantly beyond the previous bound.

Subhalo effective density slope measurements from HST strong lensing data with neural likelihood-ratio estimation
Gemma Zhang, Atınç Çağan Şengül, Cora Dvorkin
Monthly Notices of the Royal Astronomical Society, 2024, Volume 527, Issue 2 [ arXiv:2308.09739 | code ] published
Astrophysics

Abstract Examining the properties of subhalos with strong gravitational lensing images can shed light on the nature of dark matter. From upcoming large-scale surveys, we expect to discover orders of magnitude more strong lens systems that can be used for subhalo studies. To optimally extract information from a large number of strong lensing images, machine learning provides promising avenues for efficient analysis that is unachievable with traditional analysis methods, but application of machine learning techniques to real observations is still limited. We build upon previous work, which uses a neural likelihood-ratio estimator, to constrain the effective density slopes of subhalos and demonstrate the feasibility of this method on real strong lensing observations. To do this, we implement significant improvements to the forward simulation pipeline and undertake careful model evaluation using simulated images. Ultimately, we use our trained model to predict the effective subhalo density slope from combining a set of strong lensing images taken by the extit{Hubble Space Telescope}. We found the subhalo slope measurement of this set of observations to be steeper than the slope predictions of cold dark matter subhalos. Our result adds to several previous works that also measured high subhalo slopes in observations. Although a possible explanation for this is that subhalos with steeper slopes are easier to detect due to selection effects and thus contribute to statistical bias, our result nevertheless points to the need for careful analysis of more strong lensing observations from future surveys.

Data Compression and Inference in Cosmology with Self-Supervised Machine Learning
Aizhan Akhmetzhanova, Siddharth Mishra-Sharma, Cora Dvorkin
Monthly Notices of the Royal Astronomical Society, 2023, Volume 527, Issue 3 [ arXiv:2308.09751 ] published
Astrophysics

Abstract The influx of massive amounts of data from current and upcoming cosmological surveys necessitates compression schemes that can efficiently summarize the data with minimal loss of information. We introduce a method that leverages the paradigm of self-supervised machine learning in a novel manner to construct representative summaries of massive datasets using simulation-based augmentations. Deploying the method on hydrodynamical cosmological simulations, we show that it can deliver highly informative summaries, which can be used for a variety of downstream tasks, including precise and accurate parameter inference. We demonstrate how this paradigm can be used to construct summary representations that are insensitive to prescribed systematic effects, such as the influence of baryonic physics. Our results indicate that self-supervised machine learning techniques offer a promising new approach for compression of cosmological data as well its analysis.

Gravitational action for a massive Majorana fermion in 2d quantum gravity
Corinne de Lacroix, Harold Erbin, Vincent Lahoche
Journal of High Energy Physics, 2024, Volume 2024, Article number 68 [ arXiv:2308.08342 ] published
Theoretical Physics

Abstract We compute the gravitational action of a free massive Majorana fermion coupled to two-dimensional gravity on compact Riemann surfaces of arbitrary genus. The structure is similar to the case of the massive scalar. The small-mass expansion of the gravitational yields the Liouville action at zeroth order, and we can identify the Mabuchi action at first order. While the massive Majorana action is a conformal deformation of the massless Majorana CFT, we find an action different from the one given by the David-Distler-Kawai (DDK) ansatz.

An Extensive Hubble Space Telescope Study of the Offset and Host Light Distributions of Type I Superluminous Supernovae
Brian Hsu, Peter K. Blanchard, Edo Berger, Sebastian Gomez
The Astrophysical Journal 2024, Volume 961, Number 2 [ arXiv:2308.07271 ] published
Astrophysics

Abstract We present an extensive Hubble Space Telescope (HST) rest-frame ultraviolet (UV) imaging study of the locations of Type I superluminous supernovae (SLSNe) within their host galaxies. The sample includes 65 SLSNe with detected host galaxies in the redshift range z≈0.05−2. Using precise astrometric matching with SN images, we determine the distributions of physical and host-normalized offsets relative to the host centers, as well as the fractional flux distribution relative to the underlying UV light distribution. We find that the host-normalized offsets of SLSNe roughly track an exponential disk profile, but exhibit an overabundance of sources with large offsets of 1.5−4 times their host half-light radius. The SLSNe normalized offsets are systematically larger than those of long gamma-ray bursts (LGRBs), and even Type Ib/c and II SNe. Furthermore, we find that about 40\% of all SLSNe occur in the dimmest regions of their host galaxies (fractional flux of 0), in stark contrast to LGRBs and Type Ib/c and II SNe. We do not detect any significant trends in the locations of SLSNe as a function of redshift, or as a function of explosion and magnetar engine parameters inferred from modeling of their optical lights curves. The significant difference in SLSN locations compared to LGRBs (and normal core-collapse SNe) suggests that at least some of their progenitors follow a different evolutionary path. We speculate that SLSNe arise from massive runaway stars from disrupted binary systems, with velocities of ∼102 km s−1.

FLORAH: A generative model for halo assembly histories
Tri Nguyen, Chirag Modi, L. Y. Aaron Yung, Rachel S. Somerville
Monthly Notices of the Royal Astronomical Society, 2024, Volume 533, Issue 3 [ arXiv:2308.05145 | code ] published
Experimental Physics

Abstract The mass assembly history (MAH) of dark matter halos plays a crucial role in shaping the formation and evolution of galaxies. MAHs are used extensively in semi-analytic and empirical models of galaxy formation, yet current analytic methods to generate them are inaccurate and unable to capture their relationship with the halo internal structure and large-scale environment. This paper introduces FLORAH, a machine-learning framework for generating assembly histories of ensembles of dark matter halos. We train FLORAH on the assembly histories from the GUREFT and VSMDPL N-body simulations and demonstrate its ability to recover key properties such as the time evolution of mass and concentration. We obtain similar results for the galaxy stellar mass versus halo mass relation and its residuals when we run the Santa Cruz semi-analytic model on FLORAH-generated assembly histories and halo formation histories extracted from an N-body simulation. We further show that FLORAH also reproduces the dependence of clustering on properties other than mass (assembly bias), which is not captured by other analytic methods. By combining multiple networks trained on a suite of simulations with different redshift ranges and mass resolutions, we are able to construct accurate main progenitor branches (MPBs) with a wide dynamic mass range from z=0 up to an ultra-high redshift z≈20, currently far beyond that of a single N-body simulation. FLORAH is the first step towards a machine learning-based framework for planting full merger trees; this will enable the exploration of different galaxy formation scenarios with great computational efficiency at unprecedented accuracy.

First demonstration for a LArTPC-based search for intranuclear neutron-antineutron transitions and annihilation in 40Ar using the MicroBooNE detector
MicroBooNE collaboration
[ arXiv:2308.03924 ]
Experimental Physics

Abstract In this paper, we present a novel methodology to search for intranuclear neutron-antineutron transition (n→n¯) followed by annihilation within an 40Ar nucleus, using the MicroBooNE liquid argon time projection chamber (LArTPC) detector. A discovery of n→n¯ transition or increased lower limit on the lifetime of this process would either constitute physics beyond the Standard Model or greatly constrain theories of baryogenesis, respectively. The approach presented in this paper makes use of deep learning methods to select n→n¯ events based on their unique features and differentiate them from cosmogenic backgrounds. The achieved signal and background efficiencies are (70±6)\% and (0.0020±0.0003)\%, respectively. A demonstration of a search is performed with a data set corresponding to an exposure of 3.32×1026neutron-years, and where the background rate is constrained through direct measurement, assuming the presence of a negligible signal. With this approach, no excess of events over the background prediction is observed, setting a demonstrative lower bound on the n→n¯ lifetime in 40Ar of τm>1.1×1026years, and on the free n→n¯ transition time of τn−n¯>2.6×105s, each at the 90% confidence level. This analysis represents a first-ever proof-of-principle demonstration of the ability to search for this rare process in LArTPCs with high efficiency and low background.

A Parsec-Scale Galactic 3D Dust Map out to 1.25 kpc from the Sun
Gordian Edenhofer, Catherine Zucker, Philipp Frank, Andrew K. Saydjari, Joshua S. Speagle, Douglas Finkbeiner, Torsten Enßlin
Astronomy & Astrophysics, Forthcoming article, 2024, Section Interstellar and circumstellar matter [ arXiv:2308.01295 ] published
Astrophysics

Abstract High-resolution 3D maps of interstellar dust are critical for probing the underlying physics shaping the structure of the interstellar medium, and for foreground correction of astrophysical observations affected by dust. We aim to construct a new 3D map of the spatial distribution of interstellar dust extinction out to a distance of 1.25 kpc from the Sun. We leverage distance and extinction estimates to 54 million nearby stars derived from the Gaia BP/RP spectra. Using the stellar distance and extinction information, we infer the spatial distribution of dust extinction. We model the logarithmic dust extinction with a Gaussian Process in a spherical coordinate system via Iterative Charted Refinement and a correlation kernel inferred in previous work. We probe our 661 million dimensional posterior distribution using the variational inference method MGVI. Our 3D dust map achieves an angular resolution of 14' (Nside = 256). We sample the dust extinction in 516 distance bins spanning 69 pc to 1250 pc. We obtain a maximum distance resolution of 0.4 pc at 69 pc and a minimum distance resolution of 7 pc at 1.25 kpc. Our map resolves the internal structure of hundreds of molecular clouds in the solar neighborhood and will be broadly useful for studies of star formation, Galactic structure, and young stellar populations.

Distilled Feature Fields Enable Few-Shot Language-Guided Manipulation
William Shen, Ge Yang, Alan Yu, Jansen Wong, Leslie Pack Kaelbling, Phillip Isola
[ arXiv:2308.07931 | code ]
Foundational AI

Abstract Self-supervised and language-supervised image models contain rich knowledge of the world that is important for generalization. Many robotic tasks, however, require a detailed understanding of 3D geometry, which is often lacking in 2D image features. This work bridges this 2D-to-3D gap for robotic manipulation by leveraging distilled feature fields to combine accurate 3D geometry with rich semantics from 2D foundation models. We present a few-shot learning method for 6-DOF grasping and placing that harnesses these strong spatial and semantic priors to achieve in-the-wild generalization to unseen objects. Using features distilled from a vision-language model, CLIP, we present a way to designate novel objects for manipulation via free-text natural language, and demonstrate its ability to generalize to unseen expressions and novel categories of objects.

Score-based Diffusion Models for Generating Liquid Argon Time Projection Chamber Images
Zeviel Imani, Shuchin Aeron, Taritree Wongjirad
[ arXiv:2307.13687 | code ]
Theoretical Physics

Abstract We show for the first time, high-fidelity generation of LArTPC-like data using a generative neural network. This demonstrates that methods developed for natural images do transfer to LArTPC-produced images which in contrast to natural images are globally sparse, but locally dense. We present the method we employ, which is a variant of score-based generative diffusion models. We evaluate the fidelity of the generated images using several different approaches that include using a variant of measures used to evaluate natural images, comparisons between high-dimensional distributions, and comparisons relevant to LArTPC experiments.

NuCLR, Nuclear Co-Learned Representations
Ouail Kitouni, Niklas Nolte, Sokratis Trifinopoulos, Subhash Kantamneni, Mike Williams
[ arXiv:2306.06099 ]
Experimental Physics

Abstract We introduce Nuclear Co-Learned Representations (NuCLR), a deep learning model that predicts various nuclear observables, including binding and decay energies, and nuclear charge radii. The model is trained using a multi-task approach with shared representations and obtains state-of-the-art performance, achieving levels of precision that are crucial for understanding fundamental phenomena in nuclear (astro)physics. We also report an intriguing finding that the learned representation of NuCLR exhibits the prominent emergence of crucial aspects of the nuclear shell model, namely the shell structure, including the well-known magic numbers, and the Pauli Exclusion Principle. This suggests that the model is capable of capturing the underlying physical principles and that our approach has the potential to offer valuable insights into nuclear theory.

Polarization Multi-Image Synthesis with Birefringent Metasurfaces
Dean Hazineh, Soon Wei Daniel Lim, Qi Guo, Federico Capasso, Todd Zickler
IEEE International Conference on Computational Photography (ICCP), 2023 [ arXiv:2307.08106 | code ] published
Foundational AI

Abstract Optical metasurfaces composed of precisely engineered nanostructures have gained significant attention for their ability to manipulate light and implement distinct functionalities based on the properties of the incident field. Computational imaging systems have started harnessing this capability to produce sets of coded measurements that benefit certain tasks when paired with digital post-processing. Inspired by these works, we introduce a new system that uses a birefringent metasurface with a polarizer-mosaicked photosensor to capture four optically-coded measurements in a single exposure. We apply this system to the task of incoherent opto-electronic filtering, where digital spatial-filtering operations are replaced by simpler, per-pixel sums across the four polarization channels, independent of the spatial filter size. In contrast to previous work on incoherent opto-electronic filtering that can realize only one spatial filter, our approach can realize a continuous family of filters from a single capture, with filters being selected from the family by adjusting the post-capture digital summation weights. To find a metasurface that can realize a set of user-specified spatial filters, we introduce a form of gradient descent with a novel regularizer that encourages light efficiency and a high signal-to-noise ratio. We demonstrate several examples in simulation and with fabricated prototypes, including some with spatial filters that have prescribed variations with respect to depth and wavelength.

From Discovery to the First Month of the Type II Supernova 2023ixf: High and Variable Mass Loss in the Final Year before Explosion
Daichi Hiramatsu, Daichi Tsuna, Edo Berger, Koichi Itagaki, Jared A. Goldberg, Sebastian Gomez, Kishalay De, Griffin Hosseinzadeh, K. Azalee Bostroem, Peter J. Brown, Iair Arcavi, Allyson Bieryla, Peter K. Blanchard, Gilbert A. Esquerdo, Joseph Farah, D. Andrew Howell, Tatsuya Matsumoto, Curtis McCully, Megan Newsome, Estefania Padilla Gonzalez, Craig Pellegrino, Jaehyon Rhee, Giacomo Terreran, József Vinkó, J. Craig Wheeler
The Astrophysical Journal Letters 2023, Volume 955, Number 1 [ arXiv:2307.03165 ] published
Astrophysics

Abstract We present the discovery of the Type II supernova SN 2023ixf in M101 and follow-up photometric and spectroscopic observations, respectively, in the first month and week of its evolution. Our discovery was made within a day of estimated first light, and the following light curve is characterized by a rapid rise (≈5 days) to a luminous peak (MV≈−18.2 mag) and plateau (MV≈−17.6 mag) extending to 30 days with a fast decline rate of ≈0.03 mag day−1. During the rising phase, U−V color shows blueward evolution, followed by redward evolution in the plateau phase. Prominent flash features of hydrogen, helium, carbon, and nitrogen dominate the spectra up to ≈5 days after first light, with a transition to a higher ionization state in the first ≈2 days. Both the U−V color and flash ionization states suggest a rise in the temperature, indicative of a delayed shock breakout inside dense circumstellar material (CSM). From the timescales of CSM interaction, we estimate its compact radial extent of ∼(3−7)×1014 cm. We then construct numerical light-curve models based on both continuous and eruptive mass-loss scenarios shortly before explosion. For the continuous mass-loss scenario, we infer a range of mass-loss history with 0.1−1.0M⊙yr−1 in the final 2−1 yr before explosion, with a potentially decreasing mass loss of 0.01−0.1M⊙yr−1 in ∼0.7−0.4 yr toward the explosion. For the eruptive mass-loss scenario, we favor eruptions releasing 0.3−1M⊙ of the envelope at about a year before explosion, which result in CSM with mass and extent similar to the continuous scenario. We discuss the implications of the available multiwavelength constraints obtained thus far on the progenitor candidate and SN 2023ixf to our variable CSM models.

Neural Network Field Theories: Non-Gaussianity, Actions, and Locality
Mehmet Demirtas, James Halverson, Anindita Maiti, Matthew D. Schwartz, Keegan Stoner
[ arXiv:2307.03223 ]
Theoretical Physics

Abstract Both the path integral measure in field theory and ensembles of neural networks describe distributions over functions. When the central limit theorem can be applied in the infinite-width (infinite-N) limit, the ensemble of networks corresponds to a free field theory. Although an expansion in 1/N corresponds to interactions in the field theory, others, such as in a small breaking of the statistical independence of network parameters, can also lead to interacting theories. These other expansions can be advantageous over the 1/N-expansion, for example by improved behavior with respect to the universal approximation theorem. Given the connected correlators of a field theory, one can systematically reconstruct the action order-by-order in the expansion parameter, using a new Feynman diagram prescription whose vertices are the connected correlators. This method is motivated by the Edgeworth expansion and allows one to derive actions for neural network field theories. Conversely, the correspondence allows one to engineer architectures realizing a given field theory by representing action deformations as deformations of neural network parameter densities. As an example, ϕ4 theory is realized as an infinite-N neural network field theory.

The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks
Ziqian Zhong, Ziming Liu, Max Tegmark, Jacob Andreas
[ arXiv:2306.17844 | code ]
Foundational AI

Abstract Do neural networks, trained on well-understood algorithmic tasks, reliably rediscover known algorithms for solving those tasks? Several recent studies, on tasks ranging from group arithmetic to in-context linear regression, have suggested that the answer is yes. Using modular addition as a prototypical problem, we show that algorithm discovery in neural networks is sometimes more complex. Small changes to model hyperparameters and initializations can induce the discovery of qualitatively different algorithms from a fixed training set, and even parallel implementations of multiple such algorithms. Some networks trained to perform modular addition implement a familiar Clock algorithm; others implement a previously undescribed, less intuitive, but comprehensible procedure which we term the Pizza algorithm, or a variety of even more complex procedures. Our results show that even simple learning problems can admit a surprising diversity of solutions, motivating the development of new tools for characterizing the behavior of neural networks across their algorithmic phase space.

Hierarchical Neural Simulation-Based Inference Over Event Ensembles
Lukas Heinrich, Siddharth Mishra-Sharma, Chris Pollard, Philipp Windischhofer
[ arXiv:2306.12584 | code ]
Theoretical Physics

Abstract When analyzing real-world data it is common to work with event ensembles, which comprise sets of observations that collectively constrain the parameters of an underlying model of interest. Such models often have a hierarchical structure, where "local" parameters impact individual events and "global" parameters influence the entire dataset. We introduce practical approaches for optimal dataset-wide probabilistic inference in cases where the likelihood is intractable, but simulations can be realized via forward modeling. We construct neural estimators for the likelihood(-ratio) or posterior and show that explicitly accounting for the model's hierarchical structure can lead to tighter parameter constraints. We ground our discussion using case studies from the physical sciences, focusing on examples from particle physics (particle collider data) and astrophysics (strong gravitational lensing observations).

Development of the Topological Trigger for LHCb Run 3
Nicole Schulte, Blaise Raheem Delaney, Niklas Nolte, Gregory Max Ciezarek, Johannes Albrecht, Mike Williams
[ arXiv:2306.09873 ]
Experimental Physics

Abstract The data-taking conditions expected in Run 3 of the LHCb experiment at CERN are unprecedented and challenging for the software and computing systems. Despite that, the LHCb collaboration pioneers the use of a software-only trigger system to cope with the increased event rate efficiently. The beauty physics programme of LHCb is heavily reliant on topological triggers. These are devoted to selecting beauty-hadron candidates inclusively, based on the characteristic decay topology and kinematic properties expected from beauty decays. The following proceeding describes the current progress of the Run 3 implementation of the topological triggers using Lipschitz monotonic neural networks. This architecture offers robustness under varying detector conditions and sensitivity to long-lived candidates, improving the possibility of discovering New Physics at LHCb.

Quantum Computation and Simulation using Fermion-Pair Registers
Xiangkai Sun, Di Luo, Soonwon Choi
[ arXiv:2306.03905 ]
Theoretical Physics

Abstract We propose and analyze an approach to realize quantum computation and simulation using fermionic particles under quantum gas microscopes. Our work is inspired by a recent experimental demonstration of large-scale quantum registers, where tightly localized fermion pairs are used to encode qubits exhibiting long coherence time and robustness against laser intensity noise. We describe how to engineer the SWAP gate and high-fidelity controlled-phase gates by adjusting the fermion hopping as well as Feshbach interaction strengths. Combined with previously demonstrated single-qubit rotations, these gates establish the computational universality of the system. Furthermore, we show that 2D quantum Ising Hamiltonians with tunable transverse and longitudinal fields can be efficient simulated by modulating Feshbach interaction strengths. We present a sample-efficient protocol to characterize engineered gates and Hamiltonian dynamics based on an improved classical shadow process tomography that requires minimal experimental controls. Our work opens up new opportunities to harness existing ultracold quantum gases for quantum information sciences.

Discovering New Interpretable Conservation Laws as Sparse Invariants
Ziming Liu, Patrick Obin Sturm, Saketh Bharadwaj, Sam Silva, Max Tegmark
[ arXiv:2305.19525 ]
Foundational AI

Abstract Discovering conservation laws for a given dynamical system is important but challenging. In a theorist setup (differential equations and basis functions are both known), we propose the Sparse Invariant Detector (SID), an algorithm that auto-discovers conservation laws from differential equations. Its algorithmic simplicity allows robustness and interpretability of the discovered conserved quantities. We show that SID is able to rediscover known and even discover new conservation laws in a variety of systems. For two examples in fluid mechanics and atmospheric chemistry, SID discovers 14 and 3 conserved quantities, respectively, where only 12 and 2 were previously known to domain experts.

Learning Linear Groups in Neural Networks
Emmanouil Theodosis, Karim Helwani, Demba Ba
[ arXiv:2305.18552 ]
Foundational AI

Abstract Employing equivariance in neural networks leads to greater parameter efficiency and improved generalization performance through the encoding of domain knowledge in the architecture; however, the majority of existing approaches require an a priori specification of the desired symmetries. We present a neural network architecture, Linear Group Networks (LGNs), for learning linear groups acting on the weight space of neural networks. Linear groups are desirable due to their inherent interpretability, as they can be represented as finite matrices. LGNs learn groups without any supervision or knowledge of the hidden symmetries in the data and the groups can be mapped to well known operations in machine learning. We use LGNs to learn groups on multiple datasets while considering different downstream tasks; we demonstrate that the linear group structure depends on both the data distribution and the considered task.

Machine Learning for Quantum-Enhanced Gravitational-Wave Observatories
Chris Whittle, Ge Yang, Matthew Evans, Lisa Barsotti
Physical Review D, Volume 108, Article 043034 [ arXiv:2305.13780 ] published
Foundational AI

Abstract Machine learning has become an effective tool for processing the extensive data sets produced by large physics experiments. Gravitational-wave detectors are now listening to the universe with quantum-enhanced sensitivity, accomplished with the injection of squeezed vacuum states. Squeezed state preparation and injection is operationally complicated, as well as highly sensitive to environmental fluctuations and variations in the interferometer state. Achieving and maintaining optimal squeezing levels is a challenging problem and will require development of new techniques to reach the lofty targets set by design goals for future observing runs and next-generation detectors. We use machine learning techniques to predict the squeezing level during the third observing run of the Laser Interferometer Gravitational-Wave Observatory (LIGO) based on auxiliary data streams, and offer interpretations of our models to identify and quantify salient sources of squeezing degradation. The development of these techniques lays the groundwork for future efforts to optimize squeezed state injection in gravitational-wave detectors, with the goal of enabling closed-loop control of the squeezer subsystem by an agent based on machine learning.

Materialistic: Selecting Similar Materials in Images
Prafull Sharma, Julien Philip, Michael Gharbi, Bill Freeman, Fredo Durand, Valentin Deschaintre
ACM Transactions on Graphics, 2023, Volume 42, Issue 4 [ arXiv:2305.13291 ] published
Foundational AI

Abstract Separating an image into meaningful underlying components is a crucial first step for both editing and understanding images. We present a method capable of selecting the regions of a photograph exhibiting the same material as an artist-chosen area. Our proposed approach is robust to shading, specular highlights, and cast shadows, enabling selection in real images. As we do not rely on semantic segmentation (different woods or metal should not be selected together), we formulate the problem as a similarity-based grouping problem based on a user-provided image location. In particular, we propose to leverage the unsupervised DINO features coupled with a proposed Cross-Similarity module and an MLP head to extract material similarities in an image. We train our model on a new synthetic image dataset, that we release. We show that our method generalizes well to real-world images. We carefully analyze our model's behavior on varying material properties and lighting. Additionally, we evaluate it against a hand-annotated benchmark of 50 real photographs. We further demonstrate our model on a set of applications, including material editing, in-video selection, and retrieval of object photographs with similar materials.

Dynamic Sparse Training with Structured Sparsity
Mike Lasby, Anna Golubeva, Utku Evci, Mihai Nica, Yani Ioannou
[ arXiv:2305.02299 ]
Foundational AI

Abstract Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically cheaper to train, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method to learn a variant of structured N:M sparsity by imposing a constant fan-in constraint. We demonstrate with both a theoretical analysis and empirical results: state-of-the-art spare-to-sparse structured DST performance on a variety of network architectures, a condensed representation with a reduced parameter and memory footprint, and reduced inference time compared to dense models with a naive PyTorch CPU implementation of the condensed representation.

Seeing is Believing: Brain-Inspired Modular Training for Mechanistic Interpretability
Ziming Liu, Eric Gan, Max Tegmark
[ arXiv:2305.08746 ]
Foundational AI

Abstract We introduce Brain-Inspired Modular Training (BIMT), a method for making neural networks more modular and interpretable. Inspired by brains, BIMT embeds neurons in a geometric space and augments the loss function with a cost proportional to the length of each neuron connection. We demonstrate that BIMT discovers useful modular neural networks for many simple tasks, revealing compositional structures in symbolic formulas, interpretable decision boundaries and features for classification, and mathematical structure in algorithmic datasets. The ability to directly see modules with the naked eye can complement current mechanistic interpretability strategies such as probes, interventions or staring at all weights.

Multiple Peaks and a Long Precursor in the Type IIn Supernova 2021qqp: An Energetic Explosion in a Complex Circumstellar Environment
Daichi Hiramatsu, Tatsuya Matsumoto, Edo Berger, Conor Ransome, V. Ashley Villar, Sebastian Gomez, Yvette Cendes, Kishalay De, K. Azalee Bostroem, Joseph Farah, D. Andrew Howell, Curtis McCully, Megan Newsome, Estefania Padilla Gonzalez, Craig Pellegrino, Akihiro Suzuki, Giacomo Terreran
The Astrophysical Journal, 2024, Volume 964, Number 2 [ arXiv:2305.11168 ] published
Astrophysics

Abstract We present optical photometry and spectroscopy of the Type IIn supernova (SN) 2021qqp. Its unusual light curve is marked by a long precursor for ≈300 days, a rapid increase in brightness for ≈60 days, and then a sharp increase of ≈1.6 mag in only a few days to a first peak of Mr≈−19.5 mag. The light curve then declines rapidly until it re-brightens to a second distinct peak of Mr≈−17.3 mag centered at ≈335 days after the first peak. The spectra are dominated by Balmer lines with a complex morphology, including a narrow component with a width of ≈1300 km s−1 (first peak) and ≈2500 km s−1 (second peak) that we associate with the circumstellar medium (CSM) and a P Cygni component with an absorption velocity of ≈8500 km s−1 (first peak) and ≈5600 km s−1 (second peak) that we associate with the SN-CSM interaction shell. Using the luminosity and velocity evolution, we construct a flexible analytical model, finding two significant mass-loss episodes with peak mass loss rates of ≈10 and ≈5M⊙ yr−1 about 0.8 and 2 yr before explosion, respectively, with a total CSM mass of ≈2−4M⊙. We show that the most recent mass-loss episode could explain the precursor for the year preceding the explosion. The SN ejecta mass is constrained to be ≈5−30M⊙ for an explosion energy of ≈(3−10)×1051 erg. We discuss eruptive massive stars (luminous blue variable, pulsational pair instability) and an extreme stellar merger with a compact object as possible progenitor channels.

Constraint of pionless EFT using two-nucleon spectra from lattice QCD
William Detmold, Fernando Romero-López, Phiala E. Shanahan
[ arXiv:2305.06313 ]
Theoretical Physics

Abstract Finite-volume pionless effective field theory (FVEFTπ/) at next-to-leading order (NLO) is used to analyze the two-nucleon lattice QCD spectrum of Ref.~\cite{Amarasinghe:2021lqa}, performed at quark masses corresponding to a pion mass of approximately 800 MeV. Specifically, the effective theory is formulated in finite volume, and variational sets of wave functions are optimized using differential programming. Using these wave functions projected to the appropriate finite-volume symmetry group, variational bounds from FVEFTπ/ are obtained for the ground state, as well as excited states. By comparison with the lattice QCD GEVP spectrum, different low energy constants (LECs) are constrained. Relativistic corrections are incorporated, allowing for the extractions of NLO LECs, as well as the leading s-d-wave mixing term in the deuteron channel.

Symbolic Regression on FPGAs for Fast Machine Learning Inference
Ho Fung Tsoi, Adrian Alan Pol, Vladimir Loncar, Ekaterina Govorkova, Miles Cranmer, Sridhara Dasu, Peter Elmer, Philip Harris, Isobel Ojalvo, Maurizio Pierini
EPJ Web of Conferences 2024, Volume 295 [ arXiv:2305.04099 | code ] published
Experimental Physics

Abstract The high-energy physics community is investigating the feasibility of deploying machine-learning-based solutions on Field-Programmable Gate Arrays (FPGAs) to improve physics sensitivity while meeting data processing latency limitations. In this contribution, we introduce a novel end-to-end procedure that utilizes a machine learning technique called symbolic regression (SR). It searches equation space to discover algebraic relations approximating a dataset. We use PySR (software for uncovering these expressions based on evolutionary algorithm) and extend the functionality of hls4ml (a package for machine learning inference in FPGAs) to support PySR-generated expressions for resource-constrained production environments. Deep learning models often optimise the top metric by pinning the network size because vast hyperparameter space prevents extensive neural architecture search. Conversely, SR selects a set of models on the Pareto front, which allows for optimising the performance-resource tradeoff directly. By embedding symbolic forms, our implementation can dramatically reduce the computational resources needed to perform critical tasks. We validate our procedure on a physics benchmark: multiclass classification of jets produced in simulated proton-proton collisions at the CERN Large Hadron Collider, and show that we approximate a 3-layer neural network with an inference model that has as low as 5 ns execution time (a reduction by a factor of 13) and over 90% approximation accuracy.

A Spectral Metric for Collider Geometry
Andrew J. Larkoski, Jesse Thaler
Journal of High Energy Physics 2023, Volume 2023, article number 107 [ arXiv:2305.03751 ] published
Theoretical Physics

Abstract By quantifying the distance between two collider events, one can triangulate a metric space and reframe collider data analysis as computational geometry. One popular geometric approach is to first represent events as an energy flow on an idealized celestial sphere and then define the metric in terms of optimal transport in two dimensions. In this paper, we advocate for representing events in terms of a spectral function that encodes pairwise particle angles and products of particle energies, which enables a metric distance defined in terms of one-dimensional optimal transport. This approach has the advantage of automatically incorporating obvious isometries of the data, like rotations about the colliding beam axis. It also facilitates first-principles calculations, since there are simple closed-form expressions for optimal transport in one dimension. Up to isometries and event sets of measure zero, the spectral representation is unique, so the metric on the space of spectral functions is a metric on the space of events. At lowest order in perturbation theory in electron-positron collisions, our metric is simply the summed squared invariant masses of the two event hemispheres. Going to higher orders, we present predictions for the distribution of metric distances between jets in fixed-order and resummed perturbation theory as well as in parton-shower generators. Finally, we speculate on whether the spectral approach could furnish a useful metric on the space of quantum field theories.

Normalizing flows for lattice gauge theory in arbitrary space-time dimension
Ryan Abbott, Michael S. Albergo, Aleksandar Botev, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Gurtej Kanwar, Alexander G.D.G. Matthews, Sébastien Racanière, Ali Razavi, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2305.02402 | code ]
Theoretical Physics

Abstract Applications of normalizing flows to the sampling of field configurations in lattice gauge theory have so far been explored almost exclusively in two space-time dimensions. We report new algorithmic developments of gauge-equivariant flow architectures facilitating the generalization to higher-dimensional lattice geometries. Specifically, we discuss masked autoregressive transformations with tractable and unbiased Jacobian determinants, a key ingredient for scalable and asymptotically exact flow-based sampling algorithms. For concreteness, results from a proof-of-principle application to SU(3) lattice gauge theory in four space-time dimensions are reported.

Pileup and Infrared Radiation Annihilation (PIRANHA): A Paradigm for Continuous Jet Grooming
Samuel Alipour-Fard, Patrick T. Komiske, Eric M. Metodiev, Jesse Thaler
Journal of High Energy Physics 2023, Volume 2023, Article number 157 [ arXiv:2305.00989 | code ] published
Experimental Physics

Abstract Jet grooming is an important strategy for analyzing relativistic particle collisions in the presence of contaminating radiation. Most jet grooming techniques introduce hard cutoffs to remove soft radiation, leading to discontinuous behavior and associated experimental and theoretical challenges. In this paper, we introduce Pileup and Infrared Radiation Annihilation (PIRANHA), a paradigm for continuous jet grooming that overcomes the discontinuity and infrared sensitivity of hard-cutoff grooming procedures. We motivate PIRANHA from the perspective of optimal transport and the Energy Movers Distance and review Apollonius Subtraction and Iterated Voronoi Subtraction as examples of PIRANHA-style grooming. We then introduce a new tree-based implementation of PIRANHA, Recursive Subtraction, with reduced computational costs. Finally, we demonstrate the performance of Recursive Subtraction in mitigating sensitivity to soft distortions from hadronization and detector effects, and additive contamination from pileup and the underlying event.

Prometheus: An Open-Source Neutrino Telescope Simulation
Jeffrey Lazar, Stephan Meighen-Berger, Christian Haack, David Kim, Santiago Giner, Carlos A. Argüelles
[ arXiv:2304.14526 | code ]
Experimental Physics

Abstract Neutrino telescopes are gigaton-scale neutrino detectors comprised of individual light-detection units. Though constructed from simple building blocks, they have opened a new window to the Universe and are able to probe center-of-mass energies that are comparable to those of collider experiments. \prometheus{} is a new, open-source simulation tailored for this kind of detector. Our package, which is written in a combination of \texttt{C++} and \texttt{Python} provides a balance of ease of use and performance and allows the user to simulate a neutrino telescope with arbitrary geometry deployed in ice or water. \prometheus{} simulates the neutrino interactions in the volume surrounding the detector, computes the light yield of the hadronic shower and the out-going lepton, propagates the photons in the medium, and records their arrival times and position in user-defined regions. Finally, \prometheus{} events are serialized into a \texttt{parquet} file, which is a compact and interoperational file format that allows prompt access to the events for further analysis.

Searching for ribbons with machine learning
Sergei Gukov, James Halverson, Ciprian Manolescu, Fabian Ruehle
[ arXiv:2304.09304 | code ]
Theoretical Physics

Abstract We apply Bayesian optimization and reinforcement learning to a problem in topology: the question of when a knot bounds a ribbon disk. This question is relevant in an approach to disproving the four-dimensional smooth Poincaré conjecture; using our programs, we rule out many potential counterexamples to the conjecture. We also show that the programs are successful in detecting many ribbon knots in the range of up to 70 crossings.

Correlation function distributions for O(N) lattice field theories in the disordered phase
Cagin Yunus, William Detmold
[ arXiv:2304.03820 ]
Theoretical Physics

Abstract Numerical computations in strongly-interacting quantum field theories are often performed using Monte-Carlo sampling methods. A key task in these calculations is to estimate the value of a given physical quantity from the distribution of stochastic samples that are generated using the Monte-Carlo method. Typically, the sample mean and sample variance are used to define the expectation values and uncertainties of computed quantities. However, the Monte-Carlo sample distribution contains more information than these basic properties and it is useful to investigate it more generally. In this work, the exact form of the probability distributions of two-point correlation functions at zero momentum in O(N) lattice field theories in the disordered phase and in infinite volume are determined. These distributions allow for a robust investigation of the efficacy of the Monte-Carlo sampling procedure and are shown also to allow for improved estimators of the target physical quantity to be constructed. The theoretical expectations are shown to agree with numerical calculations in the O(2) model.

GenPhys: From Physical Processes to Generative Models
Ziming Liu, Di Luo, Yilun Xu, Tommi Jaakkola, Max Tegmark
[ arXiv:2304.02637 ]
Foundational AI

Abstract Since diffusion models (DM) and the more recent Poisson flow generative models (PFGM) are inspired by physical processes, it is reasonable to ask: Can physical processes offer additional new generative models? We show that the answer is yes. We introduce a general family, Generative Models from Physical Processes (GenPhys), where we translate partial differential equations (PDEs) describing physical processes to generative models. We show that generative models can be constructed from s-generative PDEs (s for smooth). GenPhys subsume the two existing generative models (DM and PFGM) and even give rise to new families of generative models, e.g., "Yukawa Generative Models" inspired from weak interactions. On the other hand, some physical processes by default do not belong to the GenPhys family, e.g., the wave equation and the Schrödinger equation, but could be made into the GenPhys family with some modifications. Our goal with GenPhys is to explore and expand the design space of generative models.

ANTN: Bridging Autoregressive Neural Networks and Tensor Networks for Quantum Many-Body Simulation
Zhuo Chen, Laker Newhouse, Eddie Chen, Di Luo, Marin Soljačić
[ arXiv:2304.01996 | code ]
Theoretical Physics

Abstract Quantum many-body physics simulation has important impacts on understanding fundamental science and has applications to quantum materials design and quantum technology. However, due to the exponentially growing size of the Hilbert space with respect to the particle number, a direct simulation is intractable. While representing quantum states with tensor networks and neural networks are the two state-of-the-art methods for approximate simulations, each has its own limitations in terms of expressivity and optimization. To address these challenges, we develop a novel architecture, Autoregressive Neural TensorNet (ANTN), which bridges tensor networks and autoregressive neural networks. We show that Autoregressive Neural TensorNet parameterizes normalized wavefunctions with exact sampling, generalizes the expressivity of tensor networks and autoregressive neural networks, and inherits a variety of symmetries from autoregressive neural networks. We demonstrate our approach on the 2D J1-J2 Heisenberg model with different systems sizes and coupling parameters, outperforming both tensor networks and autoregressive neural networks. Our work opens up new opportunities for both scientific simulations and machine learning applications.

DribbleBot: Dynamic Legged Manipulation in the Wild
Yandong Ji, Gabriel B. Margolis, Pulkit Agrawal
[ arXiv:2304.01159 | code ]
Foundational AI

Abstract DribbleBot (Dexterous Ball Manipulation with a Legged Robot) is a legged robotic system that can dribble a soccer ball under the same real-world conditions as humans (i.e., in-the-wild). We adopt the paradigm of training policies in simulation using reinforcement learning and transferring them into the real world. We overcome critical challenges of accounting for variable ball motion dynamics on different terrains and perceiving the ball using body-mounted cameras under the constraints of onboard computing. Our results provide evidence that current quadruped platforms are well-suited for studying dynamic whole-body control problems involving simultaneous locomotion and manipulation directly from sensory observations.

Neural Volumetric Memory for Visual Locomotion Control
Ruihan Yang, Ge Yang, Xiaolong Wang
[ arXiv:2304.01201 ]
Foundational AI

Abstract Legged robots have the potential to expand the reach of autonomy beyond paved roads. In this work, we consider the difficult problem of locomotion on challenging terrains using a single forward-facing depth camera. Due to the partial observability of the problem, the robot has to rely on past observations to infer the terrain currently beneath it. To solve this problem, we follow the paradigm in computer vision that explicitly models the 3D geometry of the scene and propose Neural Volumetric Memory (NVM), a geometric memory architecture that explicitly accounts for the SE(3) equivariance of the 3D world. NVM aggregates feature volumes from multiple camera views by first bringing them back to the ego-centric frame of the robot. We test the learned visual-locomotion policy on a physical robot and show that our approach, which explicitly introduces geometric priors during training, offers superior performance than more naïve methods. We also include ablation studies and show that the representations stored in the neural volumetric memory capture sufficient geometric information to reconstruct the scene.

Level Crossings, Attractor Points and Complex Multiplication
Hamza Ahmed, Fabian Ruehle
Journal of High Energy Physics, 2023, Volume 2023, Article number 164 [ arXiv:2304.00027 ] published
Theoretical Physics

Abstract We study the complex structure moduli dependence of the scalar Laplacian eigenmodes for one-parameter families of Calabi-Yau n-folds in P^{n+1}. It was previously observed that some eigenmodes get lighter while others get heavier as a function of these moduli, which leads to eigenvalue crossing. We identify the cause for this behavior for the torus. We then show that at points in a sublocus of complex structure moduli space where Laplacian eigenmodes cross, the torus has complex multiplication. We speculate that the generalization to arbitrary Calabi-Yau manifolds could be that level crossing is related to rank one attractor points. To test this, we compute the eigenmodes numerically for the quartic K3 and the quintic threefold, and match crossings to CM and attractor points in these varieties. To quantify the error of our numerical methods, we also study the dependence of the numerical spectrum on the quality of the Calabi-Yau metric approximation, the number of points sampled from the Calabi-Yau variety, the truncation of the eigenbasis, and the the distance from degeneration points in complex structure moduli space.

Noisy dynamical systems evolve error correcting codes and modularity
Trevor McCourt, Ila R. Fiete, Isaac L. Chuang
[ arXiv:2303.14448 | code ]
Foundational AI

Abstract There is an intimate connection between life (as we know it) and fault tolerance. Despite residing in the stochastic and complex physical world, biological systems execute functions that allow them to survive and thrive by maintaining their physical integrity. At the same time, biological systems are strikingly modular: parts responsible for different functions tend to be physically separate and easily distinguishable. In this work, through experiments in Boolean networks, we show that the simultaneous presence of fault tolerance and modularity in biological systems is no coincidence. Rather, it is a typical co-occurrence in dynamic systems undergoing adaptive evolution in noisy environments. From this, we deduce the principle of error correction enhanced evolvability: systems possessing error-correcting codes are more effectively improved by evolution than those without. Noise creates the evolutionary pressure to develop initial error-correcting codes, suggesting it plays a larger role in the evolution of complex structures than previously thought.

The Quantization Model of Neural Scaling
Eric J. Michaud, Ziming Liu, Uzay Girit, Max Tegmark
[ arXiv:2303.13506 | code ]
Foundational AI

Abstract We propose the Quantization Model of neural scaling laws, explaining both the observed power law dropoff of loss with model and data size, and also the sudden emergence of new capabilities with scale. We derive this model from what we call the Quantization Hypothesis, where learned network capabilities are quantized into discrete chunks (quanta). We show that when quanta are learned in order of decreasing use frequency, then a power law in use frequencies explains observed power law scaling of loss. We validate this prediction on toy datasets, then study how scaling curves decompose for large language models. Using language model internals, we auto-discover diverse model capabilities (quanta) and find tentative evidence that the distribution over corresponding subproblems in the prediction of natural text is compatible with the power law predicted from the neural scaling exponent as predicted from our theory.

Artificial intelligence for artificial materials: moiré atom
Di Luo, Aidan P. Reddy, Trithep Devakul, Liang Fu
[ arXiv:2303.08162 ]
Theoretical Physics

Abstract Moiré engineering in atomically thin van der Waals heterostructures creates artificial quantum materials with designer properties. We solve the many-body problem of interacting electrons confined to a moiré superlattice potential minimum (the moiré atom) using a 2D fermionic neural network. We show that strong Coulomb interactions in combination with the anisotropic moiré potential lead to striking ``Wigner molecule" charge density distributions observable with scanning tunneling microscopy.

Exploring the CP-violating Dashen phase in the Schwinger model with tensor networks
Lena Funcke, Karl Jansen, Stefan Kühn
Physical Review D, 2023, Volume 108, Issue 1 [ arXiv:2303.03799 ] published
Theoretical Physics

Abstract We numerically study the phase structure of the two-flavor Schwinger model with matrix product states, focusing on the (1+1)-dimensional analog of the CP-violating Dashen phase in QCD. We simulate the two-flavor Schwinger model around the point where the positive mass of one fermion flavor corresponds to the negative mass of the other fermion flavor, which is a sign-problem afflicted regime for conventional Monte Carlo techniques. Our results indicate that the model undergoes a CP-violating Dashen phase transition at this point, which manifests itself in abrupt changes of the average electric field and the analog of the pion condensate in the model. Studying the scaling of the bipartite entanglement entropy as a function of the volume, we find clear indications that this transition is not of first order.

Multi-Symmetry Ensembles: Improving Diversity and Generalization via Opposing Symmetries
Charlotte Loh, Seungwook Han, Shivchander Sudalairaj, Rumen Dangovski, Kai Xu, Florian Wenzel, Marin Soljacic, Akash Srivastava
[ arXiv:2303.02484 | code ]
Foundational AI

Abstract Deep ensembles (DE) have been successful in improving model performance by learning diverse members via the stochasticity of random initialization. While recent works have attempted to promote further diversity in DE via hyperparameters or regularizing loss functions, these methods primarily still rely on a stochastic approach to explore the hypothesis space. In this work, we present Multi-Symmetry Ensembles (MSE), a framework for constructing diverse ensembles by capturing the multiplicity of hypotheses along symmetry axes, which explore the hypothesis space beyond stochastic perturbations of model weights and hyperparameters. We leverage recent advances in contrastive representation learning to create models that separately capture opposing hypotheses of invariant and equivariant functional classes and present a simple ensembling approach to efficiently combine appropriate hypotheses for a given task. We show that MSE effectively captures the multiplicity of conflicting hypotheses that is often required in large, diverse datasets like ImageNet. As a result of their inherent diversity, MSE improves classification performance, uncertainty quantification, and generalization across a series of transfer tasks.

Via Machinae 2.0: Full-Sky, Model-Agnostic Search for Stellar Streams in Gaia DR2
David Shih, Matthew R. Buckley, Lina Necib
[ arXiv:2303.01529 ]
Astrophysics

Abstract We present an update to Via Machinae, an automated stellar stream-finding algorithm based on the deep learning anomaly detector ANODE. Via Machinae identifies stellar streams within Gaia, using only angular positions, proper motions, and photometry, without reference to a model of the Milky Way potential for orbit integration or stellar distances. This new version, Via Machinae 2.0, includes many improvements and refinements to nearly every step of the algorithm, that altogether result in more robust and visually distinct stream candidates than our original formulation. In this work, we also provide a quantitative estimate of the false positive rate of Via Machinae 2.0 by applying it to a simulated Gaia-mock catalog based on Galaxia, a smooth model of the Milky Way that does not contain substructure or stellar streams. Finally, we perform the first full-sky search for stellar streams with Via Machinae 2.0, identifying 102 streams at high significance within the Gaia Data Release 2, of which only 10 have been previously identified. While follow-up observations for further confirmation are required, taking into account the false positive rate presented in this work, we expect approximately 90 of these stream candidates to correspond to real stellar structures.

Computational Mirror Symmetry
Mehmet Demirtas, Manki Kim, Liam McAllister, Jakob Moritz, Andres Rios-Tascon
Journal of High Energy Physics, 2024, Volume 2024, Article number 184 [ arXiv:2303.00757 ] published
Theoretical Physics

Abstract We present an efficient algorithm for computing the prepotential in compactifications of type II string theory on mirror pairs of Calabi-Yau threefolds in toric varieties. Applying this method, we exhibit the first systematic computation of genus-zero Gopakumar-Vafa invariants in compact threefolds with many moduli, including examples with up to 491 vector multiplets.

Q-Flow: Generative Modeling for Differential Equations of Open Quantum Dynamics with Normalizing Flows
Owen Dugan, Peter Y. Lu, Rumen Dangovski, Di Luo, Marin Soljačić
[ arXiv:2302.12235 ]
Theoretical Physics

Abstract Studying the dynamics of open quantum systems holds the potential to enable breakthroughs both in fundamental physics and applications to quantum engineering and quantum computation. Due to the high-dimensional nature of the problem, customized deep generative neural networks have been instrumental in modeling the high-dimensional density matrix ρ, which is the key description for the dynamics of such systems. However, the complex-valued nature and normalization constraints of ρ, as well as its complicated dynamics, prohibit a seamless connection between open quantum systems and the recent advances in deep generative modeling. Here we lift that limitation by utilizing a reformulation of open quantum system dynamics to a partial differential equation (PDE) for a corresponding probability distribution Q, the Husimi Q function. Thus, we model the Q function seamlessly with off-the-shelf deep generative models such as normalizing flows. Additionally, we develop novel methods for learning normalizing flow evolution governed by high-dimensional PDEs, based on the Euler method and the application of the time-dependent variational principle. We name the resulting approach Q-Flow and demonstrate the scalability and efficiency of Q-Flow on open quantum system simulations, including the dissipative harmonic oscillator and the dissipative bosonic model. Q-Flow is superior to conventional PDE solvers and state-of-the-art physics-informed neural network solvers, especially in high-dimensional systems.

SHAPER: Can You Hear the Shape of a Jet?
Demba Ba, Akshunna S. Dogra, Rikab Gambhir, Abiy Tasissa, Jesse Thaler
Journal of High Energy Physics, 2023, Volume 2023, Article 195 [ arXiv:2302.12266 | code ] published
Theoretical Physics

Abstract The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover's Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (SHAPER), which is a general framework for defining and computing shape-based observables. SHAPER generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the SHAPER framework by performing empirical jet substructure studies using several examples of new shape-based observables.

PFGM++: Unlocking the Potential of Physics-Inspired Generative Models
Yilun Xu, Ziming Liu, Yonglong Tian, Shangyuan Tong, Max Tegmark, Tommi Jaakkola
[ arXiv:2302.04265 | code ]
Foundational AI

Abstract We introduce a new family of physics-inspired generative models termed PFGM++ that unifies diffusion models and Poisson Flow Generative Models (PFGM). These models realize generative trajectories for N dimensional data by embedding paths in N+D dimensional space while still controlling the progression with a simple scalar norm of the D additional variables. The new models reduce to PFGM when D=1 and to diffusion models when D→∞. The flexibility of choosing D allows us to trade off robustness against rigidity as increasing D results in more concentrated coupling between the data and the additional variable norms. We dispense with the biased large batch field targets used in PFGM and instead provide an unbiased perturbation-based objective similar to diffusion models. To explore different choices of D, we provide a direct alignment method for transferring well-tuned hyperparameters from diffusion models (D→∞) to any finite D values. Our experiments show that models with finite D can be superior to previous state-of-the-art diffusion models on CIFAR-10/FFHQ 64×64 datasets, with FID scores of 1.91/2.43 when D=2048/128. In addition, we demonstrate that models with smaller D exhibit improved robustness against modeling errors. Code is available at this https URL

Geometry of contact: contact planning for multi-legged robots via spin models duality
Baxi Chong, Di Luo, Tianyu Wang, Gabriel Margolis, Juntao He, Pulkit Agrawal, Marin Soljačić, Daniel I. Goldman
[ arXiv:2302.03019 ]
Theoretical Physics

Abstract Contact planning is crucial in locomoting systems.Specifically, appropriate contact planning can enable versatile behaviors (e.g., sidewinding in limbless locomotors) and facilitate speed-dependent gait transitions (e.g., walk-trot-gallop in quadrupedal locomotors). The challenges of contact planning include determining not only the sequence by which contact is made and broken between the locomotor and the environments, but also the sequence of internal shape changes (e.g., body bending and limb shoulder joint oscillation). Most state-of-art contact planning algorithms focused on conventional robots (e.g.biped and quadruped) and conventional tasks (e.g. forward locomotion), and there is a lack of study on general contact planning in multi-legged robots. In this paper, we show that using geometric mechanics framework, we can obtain the global optimal contact sequence given the internal shape changes sequence. Therefore, we simplify the contact planning problem to a graph optimization problem to identify the internal shape changes. Taking advantages of the spatio-temporal symmetry in locomotion, we map the graph optimization problem to special cases of spin models, which allows us to obtain the global optima in polynomial time. We apply our approach to develop new forward and sidewinding behaviors in a hexapod and a 12-legged centipede. We verify our predictions using numerical and robophysical models, and obtain novel and effective locomotion behaviors.

Expressive Monotonic Neural Networks
Niklas Nolte, Ouail Kitouni, Mike Williams
International Conference on Learning Representations 2023 [ ] published
Experimental Physics

Abstract The monotonic dependence of the outputs of a neural network on some of its inputs is a crucial inductive bias in many scenarios where domain knowledge dic- tates such behavior. This is especially important for interpretability and fairness considerations. In a broader context, scenarios in which monotonicity is impor- tant can be found in finance, medicine, physics, and other disciplines. It is thus desirable to build neural network architectures that implement this inductive bias provably. In this work, we propose a weight-constrained architecture with a single residual connection to achieve exact monotonic dependence in any subset of the inputs. The weight constraint scheme directly controls the Lipschitz constant of the neural network and thus provides the additional benefit of robustness. Com- pared to currently existing techniques used for monotonicity, our method is sim- pler in implementation and in theory foundations, has negligible computational overhead, is guaranteed to produce monotonic dependence, and is highly expres- sive. We show how the algorithm is used to train powerful, robust, and inter- pretable discriminators that achieve competitive performance compared to current state-of-the-art methods across various benchmarks, from social applications to the classification of the decays of subatomic particles produced at the CERN Large Hadron Collider.

Learning Silhouettes with Group Sparse Autoencoders
Emmanouil Theodosis and Demba Ba
Harvard CRISP Preprint [ ] published
Astrophysics

Abstract Sparse coding has been extensively used in neuroscience to model brain-like computation by drawing analogues between neurons’ firing activity and the nonzero elements of sparse vectors. Contemporary deep learning architectures have been used to model neural activity, inspired by signal processing algorithms; however sparse coding architectures are not able to explain the higher-order categorization that has been em- pirically observed at the neural level. In this work, we pro- pose a novel model-based architecture, termed group-sprase autoencoder, that produces sparse activity patterns in line with neural modeling, but showcases a higher-level order in its ac- tivation maps. We evaluate a dense model of our architecture on MNIST and CIFAR-10 and show that it learns dictionar- ies that resemble silhouettes of the given class, while its ac- tivations have a significantly higher level order compared to sparse architectures.

Non-perturbative strong coupling at timelike momenta
Jan Horak, Jan M. Pawlowski, Jonas Turnwald, Julian M. Urban, Nicolas Wink, Savvas Zafeiropoulos
Physical Review D 2023, Volume 107, Issue 7 [ arXiv:2301.08128 ] published
Experimental Physics

Abstract We compute the strong coupling constant of Landau gauge QCD in the full complex momentum plane, both directly and via spectral reconstruction. In particular, we consider the Taylor coupling given by the product of ghost and gluon dressing functions. Assuming spectral representations for the latter, we first show that also the coupling obeys such a representation. The subsequent spectral reconstruction of the coupling data, obtained from 2+1 flavour lattice QCD results for the ghost and gluon, is based on a probabilistic inversion of this representation using Gaussian process regression with analytically enforced asymptotics. In contradistinction, our direct calculation relies on earlier reconstruction results for the ghost and gluon spectral functions themselves, as well as data obtained in functional QCD. Apart from its relevance for studies of resonances or scattering processes, the calculation also serves as a non-trivial benchmark of our reconstruction approach. The results show remarkable agreement, testifying to the reliability of the method.

EPiC-GAN: Equivariant Point Cloud Generation for Particle Jets
Erik Buhmann, Gregor Kasieczka, Jesse Thaler
SciPost Physics, 2023, Volume 15, Issue 4 [ arXiv:2301.08128 | code ] published
Theoretical Physics

Abstract With the vast data-collecting capabilities of current and future high-energy collider experiments, there is an increasing demand for computationally efficient simulations. Generative machine learning models enable fast event generation, yet so far these approaches are largely constrained to fixed data structures and rigid detector geometries. In this paper, we introduce EPiC-GAN - equivariant point cloud generative adversarial network - which can produce point clouds of variable multiplicity. This flexible framework is based on deep sets and is well suited for simulating sprays of particles called jets. The generator and discriminator utilize multiple EPiC layers with an interpretable global latent vector. Crucially, the EPiC layers do not rely on pairwise information sharing between particles, which leads to a significant speed-up over graph- and transformer-based approaches with more complex relation diagrams. We demonstrate that EPiC-GAN scales well to large particle multiplicities and achieves high generation fidelity on benchmark jet generation tasks.

First demonstration of neural sensing and control in a kilometer-scale gravitational wave observatory
Nikhil Mukund, James Lough, Aparna Bisht, Holger Wittel, Séverin Landry Nadji, Christoph Affeldt, Fabio Bergamin, Marc Brinkmann, Volker Kringel, Harald Lück, Michael Weinert, Karsten Danzmann
Physical Review Applied, 2023, Volume 20, Issue 6 [ arXiv:2301.06221 ] published
Astrophysics

Abstract Suspended optics in gravitational wave (GW) observatories are susceptible to alignment perturbations, particularly slow drifts over time, due to variations in temperature and seismic levels. Such misalignments affect the coupling of the incident laser beam into the optical cavities, degrade both circulating power and optomechanical photon squeezing and thus decrease the astrophysical sensitivity to merging binaries. Traditional alignment techniques involve differential wavefront sensing using multiple quadrant photodiodes but are often restricted in bandwidth and are limited by the sensing noise. We present the first-ever successful implementation of neural network-based sensing and control at a gravitational wave observatory and demonstrate low-frequency control of the signal recycling mirror at the GEO 600 detector. Alignment information for three critical optics is simultaneously extracted from the interferometric dark port camera images via a CNN-LSTM network architecture and is then used for MIMO control using soft actor-critic-based deep reinforcement learning. Overall sensitivity improvement achieved using our scheme demonstrates deep learning's capabilities as a viable tool for real-time sensing and control for current and next-generation GW interferometers.

Comparing Point Cloud Strategies for Collider Event Classification
Peter Onyisi, Delon Shen, Jesse Thaler
Physical Review D, 2023, Volume 108, Issue 1 [ arXiv:2212.10659 | code ] published
Theoretical Physics

Abstract In this paper, we compare several event classification architectures defined on the point cloud representation of collider events. These approaches, which are based on the frameworks of deep sets and edge convolutions, circumvent many of the difficulties associated with traditional feature engineering. To benchmark our architectures against more traditional event classification strategies, we perform a case study involving Higgs boson decays to tau leptons. We find a 2.5 times increase in performance compared to a baseline ATLAS analysis with engineered features. Our point cloud architectures can be viewed as simplified versions of graph neural networks, where each particle in the event corresponds to a graph node. In our case study, we find the best balance of performance and computational cost for simple pairwise architectures, which are based on learned edge features.

Non-parametric Lagrangian biasing from the insights of neural nets
Xiaohan Wu, Julian B. Munoz, Daniel J. Eisenstein
Journal of Cosmology and Astroparticle Physics 2023, Volume 2023 [ arXiv:2212.08095 ] published
Astrophysics

Abstract We present a Lagrangian model of galaxy clustering bias in which we train a neural net using the local properties of the smoothed initial density field to predict the late-time mass-weighted halo field. By fitting the mass-weighted halo field in the AbacusSummit simulations at z=0.5, we find that including three coarsely spaced smoothing scales gives the best recovery of the halo power spectrum. Adding more smoothing scales may lead to 2-5% underestimation of the large-scale power and can cause the neural net to overfit. We find that the fitted halo-to-mass ratio can be well described by two directions in the original high-dimension feature space. Projecting the original features into these two principal components and re-training the neural net either reproduces the original training result, or outperforms it with a better match of the halo power spectrum. The elements of the principal components are unlikely to be assigned physical meanings, partly owing to the features being highly correlated between different smoothing scales. Our work illustrates a potential need to include multiple smoothing scales when studying galaxy bias, and this can be done easily with machine-learning methods that can take in high dimensional input feature space.

Simulating 2+1D Lattice Quantum Electrodynamics at Finite Density with Neural Flow Wavefunctions
Zhuo Chen, Di Luo, Kaiwen Hu, Bryan K. Clark
[ arXiv:2212.06835 ]
Theoretical Physics

Abstract We present a neural flow wavefunction, Gauge-Fermion FlowNet, and use it to simulate 2+1D lattice compact quantum electrodynamics with finite density dynamical fermions. The gauge field is represented by a neural network which parameterizes a discretized flow-based transformation of the amplitude while the fermionic sign structure is represented by a neural net backflow. This approach directly represents the U(1) degree of freedom without any truncation, obeys Guass's law by construction, samples autoregressively avoiding any equilibration time, and variationally simulates Gauge-Fermion systems with sign problems accurately. In this model, we investigate confinement and string breaking phenomena in different fermion density and hopping regimes. We study the phase transition from the charge crystal phase to the vacuum phase at zero density, and observe the phase seperation and the net charge penetration blocking effect under magnetic interaction at finite density. In addition, we investigate a magnetic phase transition due to the competition effect between the kinetic energy of fermions and the magnetic energy of the gauge field. With our method, we further note potential differences on the order of the phase transitions between a continuous U(1) system and one with finite truncation. Our state-of-the-art neural network approach opens up new possibilities to study different gauge theories coupled to dynamical matter in higher dimensions.

Stellar Reddening Based Extinction Maps for Cosmological Applications
Nayantara Mudur, Core Francisco Park, Douglas P Finkbeiner
The Astrophysical Journal, 2023, Volume 949, Number 2 [ arXiv:2212.04514 | code ] published
Astrophysics

Abstract Cosmological surveys must correct their observations for the reddening of extragalactic objects by Galactic dust. Existing dust maps, however, have been found to have spatial correlations with the large-scale structure of the Universe. Errors in extinction maps can propagate systematic biases into samples of dereddened extragalactic objects and into cosmological measurements such as correlation functions between foreground lenses and background objects and the primordial non-gaussianity parameter fNL. Emission-based maps are contaminated by the cosmic infrared background, while maps inferred from stellar-reddenings suffer from imperfect removal of quasars and galaxies from stellar catalogs. Thus, stellar-reddening based maps using catalogs without extragalactic objects offer a promising path to making dust maps with minimal correlations with large-scale structure. We present two high-latitude integrated extinction maps based on stellar reddenings, with a point spread function of full-width half-maximum 6.1' and 15'. We employ a strict selection of catalog objects to filter out galaxies and quasars and measure the spatial correlation of our extinction maps with extragalactic structure. Our galactic extinction maps have reduced spatial correlation with large scale structure relative to most existing stellar-reddening based and emission-based extinction maps.

Measuring the 8621 Å Diffuse Interstellar Band in Gaia DR3 RVS Spectra: Obtaining a Clean Catalog by Marginalizing over Stellar Types
Andrew K. Saydjari, Catherine Zucker, J. E. G. Peek, Douglas P. Finkbeiner
The Astrophysical Journal, 954 141, 2023 [ arXiv:2212.03879 | code ] published
Astrophysics

Abstract Diffuse interstellar bands (DIBs) are broad absorption features associated with interstellar dust and can serve as chemical and kinematic tracers. Conventional measurements of DIBs in stellar spectra are complicated by residuals between observations and best-fit stellar models. To overcome this, we simultaneously model the spectrum as a combination of stellar, dust, and residual components, with full posteriors on the joint distribution of the components. This decomposition is obtained by modeling each component as a draw from a high-dimensional Gaussian distribution in the data-space (the observed spectrum) -- a method we call "Marginalized Analytic Data-space Gaussian Inference for Component Separation" (MADGICS). We use a data-driven prior for the stellar component, which avoids missing stellar features not included in synthetic line lists. This technique provides statistically rigorous uncertainties and detection thresholds, which are required to work in the low signal-to-noise regime that is commonplace for dusty lines of sight. We reprocess all public Gaia DR3 RVS spectra and present an improved 8621 Å DIB catalog, free of detectable stellar line contamination. We constrain the rest-frame wavelength to 8623.14±0.087 Å (vacuum), find no significant evidence for DIBs in the Local Bubble from the 1/6th of RVS spectra that are public, and show unprecedented correlation with kinematic substructure in Galactic CO maps. We validate the catalog, its reported uncertainties, and biases using synthetic injection tests. We believe MADGICS provides a viable path forward for large-scale spectral line measurements in the presence of complex spectral contamination.

Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior
Gabriel B Margolis, Pulkit Agrawal
[ arXiv:2212.03238 | code ]
Foundational AI

Abstract Learned locomotion policies can rapidly adapt to diverse environments similar to those experienced during training but lack a mechanism for fast tuning when they fail in an out-of-distribution test environment. This necessitates a slow and iterative cycle of reward and environment redesign to achieve good performance on a new task. As an alternative, we propose learning a single policy that encodes a structured family of locomotion strategies that solve training tasks in different ways, resulting in Multiplicity of Behavior (MoB). Different strategies generalize differently and can be chosen in real-time for new tasks or environments, bypassing the need for time-consuming retraining. We release a fast, robust open-source MoB locomotion controller, Walk These Ways, that can execute diverse gaits with variable footswing, posture, and speed, unlocking diverse downstream tasks: crouching, hopping, high-speed running, stair traversal, bracing against shoves, rhythmic dance, and more.

Yang-Mills glueball masses from spectral reconstruction
Jan M. Pawlowski, Coralie S. Schneider, Jonas Turnwald, Julian M. Urban, Nicolas Wink
Physical Review D 2023, Volume 108, Issue 7 [ arXiv:2212.01113 ] published
Theoretical Physics

Abstract We compute masses of the two lightest glueballs from spectral reconstructions of timelike interaction channels of the four-gluon vertex in Landau gauge Yang-Mills theory. The Euclidean spacelike dressings of the vertex are calculated with the functional renormalisation group. For the spectral reconstruction of these Euclidean data, we employ Gaussian process regression. The glueball resonances can be identified straightforwardly and we obtain msc=1870(75) MeV as well as mps=2700(120) MeV, in accordance with functional bound state and lattice calculations.

Variational Neural-Network Ansatz for Continuum Quantum Field Theory
John M. Martyn, Khadijeh Najafi, Di Luo
APS Journals 2023, Volume 131, Issue 8 [ arXiv:2212.00782 | code ] published
Experimental Physics

Abstract Physicists dating back to Feynman have lamented the difficulties of applying the variational principle to quantum field theories. In non-relativistic quantum field theories, the challenge is to parameterize and optimize over the infinitely many n-particle wave functions comprising the state's Fock space representation. Here we approach this problem by introducing neural-network quantum field states, a deep learning ansatz that enables application of the variational principle to non-relativistic quantum field theories in the continuum. Our ansatz uses the Deep Sets neural network architecture to simultaneously parameterize all of the n-particle wave functions comprising a quantum field state. We employ our ansatz to approximate ground states of various field theories, including an inhomogeneous system and a system with long-range interactions, thus demonstrating a powerful new tool for probing quantum field theories.

Search for boosted Higgs boson decay to a charm quark-antiquark pair in proton-proton collisions at s√ = 13 TeV
CMS Collaboration
Physical Review Letters, 2023, Volume 131, Issue 4 [ arXiv:2211.14181 ] published
Experimental Physics

Abstract A search for the standard model (SM) Higgs boson (H) produced with transverse momentum greater than 450 GeV and decaying to a charm quark-antiquark (cc¯) pair is presented. The search is performed using proton-proton collision data collected at s√ = 13 TeV by the CMS experiment at the LHC, corresponding to an integrated luminosity of 138 fb−1. Boosted H→cc¯ decay products are reconstructed as a single large-radius jet and identified using a deep neural network charm tagging technique. The method is validated by measurement of the Z→cc¯ decay process, which is observed with a signal strength of 1.00+0.17−0.14 (syst) ± 0.08 (theo) ± 0.06 (stat), defined as the ratio of the observed process rate to the standard model expectation. The observed (expected) upper limit on σ(H)(H→cc¯) is set at 47 (39) times the SM prediction at 95% confidence level.

Learning Integrable Dynamics with Action-Angle Networks
Ameya Daigavane, Arthur Kosmala, Miles Cranmer, Tess Smidt, Shirley Ho
[ arXiv:2211.15338 | code ]
Foundational AI

Abstract Machine learning has become increasingly popular for efficiently modelling the dynamics of complex physical systems, demonstrating a capability to learn effective models for dynamics which ignore redundant degrees of freedom. Learned simulators typically predict the evolution of the system in a step-by-step manner with numerical integration techniques. However, such models often suffer from instability over long roll-outs due to the accumulation of both estimation and integration error at each prediction step. Here, we propose an alternative construction for learned physical simulators that are inspired by the concept of action-angle coordinates from classical mechanics for describing integrable systems. We propose Action-Angle Networks, which learn a nonlinear transformation from input coordinates to the action-angle space, where evolution of the system is linear. Unlike traditional learned simulators, Action-Angle Networks do not employ any higher-order numerical integration methods, making them extremely efficient at modelling the dynamics of integrable physical systems.

Can denoising diffusion probabilistic models generate realistic astrophysical fields?
Nayantara Mudur, Douglas P. Finkbeiner
[ arXiv:2211.12444 | code ]
Astrophysics

Abstract Score-based generative models have emerged as alternatives to generative adversarial networks (GANs) and normalizing flows for tasks involving learning and sampling from complex image distributions. In this work we investigate the ability of these models to generate fields in two astrophysical contexts: dark matter mass density fields from cosmological simulations and images of interstellar dust. We examine the fidelity of the sampled cosmological fields relative to the true fields using three different metrics, and identify potential issues to address. We demonstrate a proof-of-concept application of the model trained on dust in denoising dust images. To our knowledge, this is the first application of this class of models to the interstellar medium.

Visual Dexterity: In-Hand Reorientation of Novel and Complex Object Shapes
Tao Chen, Megha Tippur, Siyang Wu, Vikash Kumar, Edward Adelson, Pulkit Agrawal
Science Robotics, 2023, Volume 8, Issue 84 [ arXiv:2211.11744 | code ] published
Foundational AI

Abstract In-hand object reorientation is necessary for performing many dexterous manipulation tasks, such as tool use in less structured environments that remain beyond the reach of current robots. Prior works built reorientation systems assuming one or many of the following: reorienting only specific objects with simple shapes, limited range of reorientation, slow or quasistatic manipulation, simulation-only results, the need for specialized and costly sensor suites, and other constraints which make the system infeasible for real-world deployment. We present a general object reorientation controller that does not make these assumptions. It uses readings from a single commodity depth camera to dynamically reorient complex and new object shapes by any rotation in real-time, with the median reorientation time being close to seven seconds. The controller is trained using reinforcement learning in simulation and evaluated in the real world on new object shapes not used for training, including the most challenging scenario of reorienting objects held in the air by a downward-facing hand that must counteract gravity during reorientation. Our hardware platform only uses open-source components that cost less than five thousand dollars. Although we demonstrate the ability to overcome assumptions in prior work, there is ample scope for improving absolute performance. For instance, the challenging duck-shaped object not used for training was dropped in 56 percent of the trials. When it was not dropped, our controller reoriented the object within 0.4 radians (23 degrees) 75 percent of the time. Videos are available at: this https URL.

Characterizing 4-string contact interaction using machine learning
Harold Erbin, Atakan Hilmi Fırat
Journal of High Energy Physics, 2024, Article 16 [ arXiv:2211.09129 | code ] published
Theoretical Physics

Abstract The geometry of 4-string contact interaction of closed string field theory is characterized using machine learning. We obtain Strebel quadratic differentials on 4-punctured spheres as a neural network by performing unsupervised learning with a custom-built loss function. This allows us to solve for local coordinates and compute their associated mapping radii numerically. We also train a neural network distinguishing vertex from Feynman region. As a check, 4-tachyon contact term in the tachyon potential is computed and a good agreement with the results in the literature is observed. We argue that our algorithm is manifestly independent of number of punctures and scaling it to characterize the geometry of n-string contact interaction is feasible.

Aspects of scaling and scalability for flow-based sampling of lattice QCD
Ryan Abbott, Michael S. Albergo, Aleksandar Botev, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Alexander G. D. G. Matthews, Sébastien Racanière, Ali Razavi, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
The European Physical Journal A 2023, Volume 59, Article Number 257 [ arXiv:2211.07541 ] published
Theoretical Physics

Abstract Recent applications of machine-learned normalizing flows to sampling in lattice field theory suggest that such methods may be able to mitigate critical slowing down and topological freezing. However, these demonstrations have been at the scale of toy models, and it remains to be determined whether they can be applied to state-of-the-art lattice quantum chromodynamics calculations. Assessing the viability of sampling algorithms for lattice field theory at scale has traditionally been accomplished using simple cost scaling laws, but as we discuss in this work, their utility is limited for flow-based approaches. We conclude that flow-based approaches to sampling are better thought of as a broad family of algorithms with different scaling properties, and that scalability must be assessed experimentally.

Limits on Simultaneous and Delayed Optical Emission from Well-localized Fast Radio Bursts
Daichi Hiramatsu, Edo Berger, Brian D. Metzger, Sebastian Gomez, Allyson Bieryla, Iair Arcavi, D. Andrew Howell, Ryan Mckinven, Nozomu Tominaga
The Astrophysical Journal Letters 2023, volume 947, number 2 [ arXiv:2211.03974 ] published
Astrophysics

Abstract We present the largest compilation to date of optical observations during and following fast radio bursts (FRBs). The data set includes our dedicated simultaneous and follow-up observations, as well as serendipitous archival survey observations, for a sample of 15 well-localized FRBs: eight repeating and seven one-off sources. Our simultaneous (and nearly simultaneous with a 0.4 s delay) optical observations of 13 (1) bursts from the repeating FRB 20220912A provide the deepest such limits to date for any extragalactic FRB, reaching a luminosity limit of νLν≲1042 erg s−1 (≲2×1041 erg s−1) with 15−400 s exposures; an optical-flux-to-radio-fluence ratio of fopt/Fradio≲10−7 ms−1 (≲10−8 ms−1); and flux ratio of fopt/fradio≲0.02−≲2×10−5 (≲10−6) on millisecond to second timescales. These simultaneous limits provide useful constraints in the context of FRB emission models, such as the pulsar magnetosphere and pulsar nebula models. Interpreting all available optical limits in the context of the synchrotron maser model, we find that they constrain the flare energies to ≲1043−1049 erg (depending on the distances of the various repeating FRBs, with ≲1039 erg for the Galactic SGR 1935+2154). These limits are generally at least an order of magnitude larger than those inferred from the FRBs themselves, although in the case of FRB 20220912A our simultaneous and rapid follow-up observations severely restrict the model parameter space. We conclude by exploring the potential of future simultaneous and rapid-response observations with large optical telescopes.

Gauge Equivariant Neural Networks for 2+1D U(1) Gauge Theory Simulations in Hamiltonian Formulation
Di Luo, Shunyue Yuan, James Stokes, Bryan K. Clark
[ arXiv:2211.03198 ]
Theoretical Physics

Abstract Gauge Theory plays a crucial role in many areas in science, including high energy physics, condensed matter physics and quantum information science. In quantum simulations of lattice gauge theory, an important step is to construct a wave function that obeys gauge symmetry. In this paper, we have developed gauge equivariant neural network wave function techniques for simulating continuous-variable quantum lattice gauge theories in the Hamiltonian formulation. We have applied the gauge equivariant neural network approach to find the ground state of 2+1-dimensional lattice gauge theory with U(1) gauge group using variational Monte Carlo. We have benchmarked our approach against the state-of-the-art complex Gaussian wave functions, demonstrating improved performance in the strong coupling regime and comparable results in the weak coupling regime.

QuACK: Accelerating Gradient-Based Quantum Optimization with Koopman Operator Learning
Di Luo, Jiayu Shen, Rumen Dangovski, Marin Soljačić
NeurIPS 2022, AI4Science Workshop [ arXiv:2211.01365 | code ] published
Theoretical Physics

Abstract Finding efficient optimization methods plays an important role for quantum optimization and quantum machine learning on near-term quantum computers. While backpropagation on classical computers is computationally efficient, obtaining gradients on quantum computers is not, because the computational complexity usually scales with the number of parameters and measurements. In this paper, we connect Koopman operator theory, which has been successful in predicting nonlinear dynamics, with natural gradient methods in quantum optimization. We propose a data-driven approach using Koopman operator learning to accelerate quantum optimization and quantum machine learning. We develop two new families of methods: the sliding window dynamic mode decomposition (DMD) and the neural DMD for efficiently updating parameters on quantum computers. We show that our methods can predict gradient dynamics on quantum computers and accelerate the variational quantum eigensolver used in quantum optimization, as well as quantum machine learning. We further implement our Koopman operator learning algorithm on a real IBM quantum computer and demonstrate their practical effectiveness.

A Solvable Model of Neural Scaling Laws
Alexander Maloney, Daniel A. Roberts, James Sully
[ arXiv:2210.16859 ]
Foundational AI

Abstract Large language models with a huge number of parameters, when trained on near internet-sized number of tokens, have been empirically shown to obey neural scaling laws: specifically, their performance behaves predictably as a power law in either parameters or dataset size until bottlenecked by the other resource. To understand this better, we first identify the necessary properties allowing such scaling laws to arise and then propose a statistical model -- a joint generative data model and random feature model -- that captures this neural scaling phenomenology. By solving this model in the dual limit of large training set size and large number of parameters, we gain insight into (i) the statistical structure of datasets and tasks that lead to scaling laws, (ii) the way nonlinear feature maps, such as those provided by neural networks, enable scaling laws when trained on these datasets, (iii) the optimality of the equiparameterization scaling of training sets and parameters, and (iv) whether such scaling laws can break down and how they behave when they do. Key findings are the manner in which the power laws that occur in the statistics of natural datasets are extended by nonlinear random feature maps and then translated into power-law scalings of the test loss and how the finite extent of the data's spectral power law causes the model's performance to plateau.

Deep Learning Detection and Classification of Gravitational Waves from Neutron Star-Black Hole Mergers
Richard Qiu, Plamen Krastev, Kiranjyot Gill, Edo Berger
Physics Letters B, 2023, Volume 840 [ arXiv:2210.15888 ] published
Astrophysics

Abstract The Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo Interferometer Collaborations have now detected all three classes of compact binary mergers: binary black hole (BBH), binary neutron star (BNS), and neutron star-black hole (NSBH). For coalescences involving neutron stars, the simultaneous observation of gravitational and electromagnetic radiation produced by an event, has broader potential to enhance our understanding of these events, and also to probe the equation of state (EOS) of dense matter. However, electromagnetic follow-up to gravitational wave (GW) events requires rapid real-time detection and classification of GW signals, and conventional detection approaches are computationally prohibitive for the anticipated rate of detection of next-generation GW detectors. In this work, we present the first deep learning based results of classification of GW signals from NSBH mergers in extit{real} LIGO data. We show for the first time that a deep neural network can successfully distinguish all three classes of compact binary mergers and separate them from detector noise. Specifically, we train a convolutional neural network (CNN) on ∼500,000 data samples of real LIGO noise with injected BBH, BNS, and NSBH GW signals, and we show that our network has high sensitivity and accuracy. Most importantly, we successfully recover the two confirmed NSBH events to-date (GW200105 and GW200115) and the two confirmed BNS mergers to-date (GW170817 and GW190425), together with ≈90% of all BBH candidate events from the third Gravitational Wave Transient Catalog, GWTC-3. These results are an important step towards low-latency real-time GW detection, enabling multi-messenger astronomy.

Large-time correlation functions in bosonic lattice field theories
Cagin Yunus, William Detmold
Physics Letter B, 2023, Volume 840, 137890 [ arXiv:2210.15789 ] published
Theoretical Physics

Abstract Large-time correlation functions have a pivotal role in extracting particle masses from Euclidean lattice field theory calculations, however little is known about the statistical properties of these quantities. In this work, the asymptotic form of the distributions of the correlation functions at vanishing momentum is determined for bosonic interacting lattice field theories with a unique gapped vacuum. It is demonstrated that the deviations from the asymptotic form at large Euclidean times can be utilized to determine the spectrum of the theory.

Precision Machine Learning
Eric J. Michaud, Ziming Liu, Max Tegmark
Entropy, 2023, 25(1) [ arXiv:2210.13447 | code ] published
Foundational AI

Abstract We explore unique considerations involved in fitting ML models to data with very high precision, as is often required for science applications. We empirically compare various function approximation methods and study how they scale with increasing parameters and data. We find that neural networks can often outperform classical approximation methods on high-dimensional examples, by auto-discovering and exploiting modular structures therein. However, neural networks trained with common optimizers are less powerful for low-dimensional cases, which motivates us to study the unique properties of neural network loss landscapes and the corresponding optimization challenges that arise in the high precision regime. To address the optimization issue in low dimensions, we develop training tricks which enable us to train neural networks to extremely low loss, close to the limits allowed by numerical precision.

Identifying Tidal Disruption Events with an Expansion of the FLEET Machine Learning Algorithm
Sebastian Gomez, V. Ashley Villar, Edo Berger, Suvi Gezari, Sjoert van Velzen, Matt Nicholl, Peter K. Blanchard, Kate. D. Alexander
The Astrophysical Journal, 2023, Volume 949, Issue 113 [ arXiv:2210.10810 ] published
Astrophysics

Abstract We present an expansion of FLEET, a machine learning algorithm optimized to select transients that are most likely to be tidal disruption events (TDEs). FLEET is based on a random forest algorithm trained on the light curves and host galaxy information of 4,779 spectroscopically classified transients. For transients with a probability of being a TDE, \ptde>0.5, we can successfully recover TDEs with a ≈40\% completeness and a ≈30\% purity when using the first 20 days of photometry, or a similar completeness and ≈50\% purity when including 40 days of photometry. We find that the most relevant features for differentiating TDEs from other transients are the normalized host separation, and the light curve (g−r) color during peak. Additionally, we use FLEET to produce a list of the 39 most likely TDE candidates discovered by the Zwicky Transient Facility that remain currently unclassified. We explore the use of FLEET for the Legacy Survey of Space and Time on the Vera C. Rubin Observatory (\textit{Rubin}) and the \textit{Nancy Grace Roman Space Telescope} (\textit{Roman}). We simulate the \textit{Rubin} and \textit{Roman} survey strategies and estimate that ∼104 TDEs could be discovered every year by \textit{Rubin}, and ∼200 TDEs per year by \textit{Roman}. Finally, we run FLEET on the TDEs in our \textit{Rubin} survey simulation and find that we can recover ∼30\% of those at a redshift z<0.5 with \ptde>0.5. This translates to ∼3,000 TDEs per year that FLEET could uncover from \textit{Rubin}. FLEET is provided as a open source package on GitHub this https URL

The First Two Years of FLEET: an Active Search for Superluminous Supernovae
Sebastian Gomez, Edo Berger, Peter K. Blanchard, Griffin Hosseinzadeh, Matt Nicholl, Daichi Hiramatsu, V. Ashley Villar, Yao Yin
The Astrophysical Journal, 2023, Volume 949, Issue 114 [ arXiv:2210.10811 | code ] published
Astrophysics

Abstract In November 2019 we began operating FLEET (Finding Luminous and Exotic Extragalactic Transients), a machine learning algorithm designed to photometrically identify Type I superluminous supernovae (SLSNe) in transient alert streams. Using FLEET, we spectroscopically classified 21 of the 50 SLSNe identified worldwide between November 2019 and January 2022. Based on our original algorithm, we anticipated that FLEET would achieve a purity of about 50\% for transients with a probability of being a SLSN, \pslsn>0.5; the true on-sky purity we obtained is closer to 80\%. Similarly, we anticipated FLEET could reach a completeness of about 30\%, and we indeed measure an upper limit on the completeness of ≈33\%. Here, we present FLEET 2.0, an updated version of FLEET trained on 4,780 transients (almost 3 times more than in FLEET 1.0). FLEET 2.0 has a similar predicted purity to FLEET 1.0, but outperforms FLEET 1.0 in terms of completeness, which is now closer to ≈40\% for transients with \pslsn>0.5. Additionally, we explore possible systematics that might arise from the use of FLEET for target selection. We find that the population of SLSNe recovered by FLEET is mostly indistinguishable from the overall SLSN population, in terms of physical and most observational parameters. We provide FLEET as an open source package on GitHub this https URL

Symmetries of Calabi-Yau Prepotentials with Isomorphic Flops
Andre Lukas, Fabian Ruehle
Journal of High Energy 2023, Article 175 [ arXiv:2210.09369 ] published
Theoretical Physics

Abstract Calabi-Yau threefolds with infinitely many flops to isomorphic manifolds have an extended Kahler cone made up from an infinite number of individual Kahler cones. These cones are related by reflection symmetries across flop walls. We study the implications of this cone structure for mirror symmetry, by considering the instanton part of the prepotential in Calabi-Yau threefolds. We show that such isomorphic flops across facets of the Kahler cone boundary give rise to symmetry groups isomorphic to Coxeter groups. In the dual Mori cone, non-flopping curve classes that are identified under these groups have the same Gopakumar-Vafa invariants. This leads to instanton prepotentials invariant under Coxeter groups, which we make manifest by introducing appropriate invariant functions. For some cases, these functions can be expressed in terms of theta functions whose appearance can be linked to an elliptic fibration structure of the Calabi-Yau manifold.

Electric-Magnetic Duality in a Class of G2-Compactifications of M-theory
James Halverson, Benjamin Sung, Jiahua Tian
Journal of High Energy Physics, 2023, Volume 2023, Article 89 [ arXiv:2210.08628 ] published
Theoretical Physics

Abstract We study electric-magnetic duality in compactifications of M-theory on twisted connected sum (TCS) G2 manifolds via duality with F-theory. Specifically, we study the physics of the D3-branes in F-theory compactified on a Calabi-Yau fourfold Y, dual to a compactification of M-theory on a TCS G2 manifold X. =2 supersymmetry is restored in an appropriate geometric limit. In that limit, we demonstrate that the dual of D3-branes probing seven-branes corresponds to the shrinking of certain surfaces and curves, yielding light particles that may carry both electric and magnetic charges. We provide evidence that the Minahan-Nemeschansky theories with En flavor symmetry may be realized in this way. The SL(2,ℤ) monodromy of the 3/7-brane system is dual to a Fourier-Mukai transform of the dual IIA/M-theory geometry in this limit, and we extrapolate this monodromy action to the global compactification. Away from the limit, the theory is broken to =1 supersymmetry by a D-term.

Learning to Optimize Quasi-Newton Methods
Isaac Liao, Rumen R. Dangovski, Jakob N. Foerster, Marin Soljačić
Transactions on Machine Learning Research, 2023 [ arXiv:2210.06171 ] published
Theoretical Physics

Abstract Fast gradient-based optimization algorithms have become increasingly essential for the computationally efficient training of machine learning models. One technique is to multiply the gradient by a preconditioner matrix to produce a step, but it is unclear what the best preconditioner matrix is. This paper introduces a novel machine learning optimizer called LODO, which tries to online meta-learn the best preconditioner during optimization. Specifically, our optimizer merges Learning to Optimize (L2O) techniques with quasi-Newton methods to learn preconditioners parameterized as neural networks; they are more flexible than preconditioners in other quasi-Newton methods. Unlike other L2O methods, LODO does not require any meta-training on a training task distribution, and instead learns to optimize on the fly while optimizing on the test task, adapting to the local characteristics of the loss landscape while traversing it. Theoretically, we show that our optimizer approximates the inverse Hessian in noisy loss landscapes and is capable of representing a wide range of inverse Hessians. We experimentally verify that our algorithm can optimize in noisy settings, and show that simpler alternatives for representing the inverse Hessians worsen performance. Lastly, we use our optimizer to train a semi-realistic deep neural network with 95k parameters at speeds comparable to those of standard neural network optimizers.

On the Importance of Calibration in Semi-supervised Learning
Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, Seungwook Han, Ligong Han, Leonid Karlinsky, Marin Soljacic, Akash Srivastava
[ arXiv:2210.04783 ]
Theoretical Physics

Abstract State-of-the-art (SOTA) semi-supervised learning (SSL) methods have been highly successful in leveraging a mix of labeled and unlabeled data by combining techniques of consistency regularization and pseudo-labeling. During pseudo-labeling, the model's predictions on unlabeled data are used for training and thus, model calibration is important in mitigating confirmation bias. Yet, many SOTA methods are optimized for model performance, with little focus directed to improve model calibration. In this work, we empirically demonstrate that model calibration is strongly correlated with model performance and propose to improve calibration via approximate Bayesian techniques. We introduce a family of new SSL models that optimizes for calibration and demonstrate their effectiveness across standard vision benchmarks of CIFAR-10, CIFAR-100 and ImageNet, giving up to 15.9% improvement in test accuracy. Furthermore, we also demonstrate their effectiveness in additional realistic and challenging problems, such as class-imbalanced datasets and in photonics science.

Omnigrok: Grokking Beyond Algorithmic Data
Ziming Liu, Eric J. Michaud, Max Tegmark
[ arXiv:2210.01117 | code ]
Foundational AI

Abstract Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the "LU mechanism" because training and test losses (against model weight norm) typically resemble "L" and "U", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning.

AI-Assisted Discovery of Quantitative and Formal Models in Social Science
Julia Balla, Sihao Huang, Owen Dugan, Rumen Dangovski, Marin Soljacic
Humanities and Social Sciences Communications [ arXiv:2210.00563 | code ] published
Foundational AI

Abstract In social science, formal and quantitative models, such as ones describing economic growth and collective action, are used to formulate mechanistic explanations, provide predictions, and uncover questions about observed phenomena. Here, we demonstrate the use of a machine learning system to aid the discovery of symbolic models that capture nonlinear and dynamical relationships in social science datasets. By extending neuro-symbolic methods to find compact functions and differential equations in noisy and longitudinal data, we show that our system can be used to discover interpretable models from real-world data in economics and sociology. Augmenting existing workflows with symbolic regression can help uncover novel relationships and explore counterfactual models during the scientific process. We propose that this AI-assisted framework can bridge parametric and non-parametric models commonly employed in social science research by systematically exploring the space of nonlinear models and enabling fine-grained control over expressivity and interpretability.

Finding NEEMo: Geometric Fitting using Neural Estimation of the Energy Mover’s Distance
Ouail Kitouni, Niklas Nolte, Mike Williams
[ arXiv:2209.15624 | code ]
Experimental Physics

Abstract A novel neural architecture was recently developed that enforces an exact upper bound on the Lipschitz constant of the model by constraining the norm of its weights in a minimal way, resulting in higher expressiveness compared to other techniques. We present a new and interesting direction for this architecture: estimation of the Wasserstein metric (Earth Mover's Distance) in optimal transport by employing the Kantorovich-Rubinstein duality to enable its use in geometric fitting applications. Specifically, we focus on the field of high-energy particle physics, where it has been shown that a metric for the space of particle-collider events can be defined based on the Wasserstein metric, referred to as the Energy Mover's Distance (EMD). This metrization has the potential to revolutionize data-driven collider phenomenology. The work presented here represents a major step towards realizing this goal by providing a differentiable way of directly calculating the EMD. We show how the flexibility that our approach enables can be used to develop novel clustering algorithms.

Poisson Flow Generative Models
Yilun Xu, Ziming Liu, Max Tegmark, Tommi Jaakkola
[ arXiv:2209.11178 | code ]
Foundational AI

Abstract We propose a new "Poisson flow" generative model (PFGM) that maps a uniform distribution on a high-dimensional hemisphere into any data distribution. We interpret the data points as electrical charges on the z=0 hyperplane in a space augmented with an additional dimension z, generating a high-dimensional electric field (the gradient of the solution to Poisson equation). We prove that if these charges flow upward along electric field lines, their initial distribution in the z=0 plane transforms into a distribution on the hemisphere of radius r that becomes uniform in the r→∞ limit. To learn the bijective transformation, we estimate the normalized field in the augmented space. For sampling, we devise a backward ODE that is anchored by the physically meaningful additional dimension: the samples hit the unaugmented data manifold when the z reaches zero. Experimentally, PFGM achieves current state-of-the-art performance among the normalizing flow models on CIFAR-10, with an Inception score of 9.68 and a FID score of 2.48. It also performs on par with the state-of-the-art SDE approaches while offering 10× to 20× acceleration on image generation tasks. Additionally, PFGM appears more tolerant of estimation errors on a weaker network architecture and robust to the step size in the Euler method.

Degeneracy Engineering for Classical and Quantum Annealing: A Case Study of Sparse Linear Regression in Collider Physics
Eric R. Anschuetz, Lena Funcke, Patrick T. Komiske, Serhii Kryhin, Jesse Thaler
Physical Review D, Volume 106, Article 056008 [ arXiv:2205.10375 ] published
Theoretical Physics

Abstract Classical and quantum annealing are computing paradigms that have been proposed to solve a wide range of optimization problems. In this paper, we aim to enhance the performance of annealing algorithms by introducing the technique of degeneracy engineering, through which the relative degeneracy of the ground state is increased by modifying a subset of terms in the objective Hamiltonian. We illustrate this novel approach by applying it to the example of ℓ0-norm regularization for sparse linear regression, which is in general an NP-hard optimization problem. Specifically, we show how to cast ℓ0-norm regularization as a quadratic unconstrained binary optimization (QUBO) problem, suitable for implementation on annealing platforms. As a case study, we apply this QUBO formulation to energy flow polynomials in high-energy collider physics, finding that degeneracy engineering substantially improves the annealing performance. Our results motivate the application of degeneracy engineering to a variety of regularized optimization problems.

Discovering Conservation Laws using Optimal Transport and Manifold Learning
Peter Y. Lu, Rumen Dangovski, Marin Soljačić
Nature Communications [ arXiv:2208.14995 | code ] published
Theoretical Physics

Abstract Conservation laws are key theoretical and practical tools for understanding, characterizing, and modeling nonlinear dynamical systems. However, for many complex dynamical systems, the corresponding conserved quantities are difficult to identify, making it hard to analyze their dynamics and build efficient, stable predictive models. Current approaches for discovering conservation laws often depend on detailed dynamical information, such as the equation of motion or fine-grained time measurements, with many recent proposals also relying on black box parametric deep learning methods. We instead reformulate this task as a manifold learning problem and propose a non-parametric approach, combining the Wasserstein metric from optimal transport with diffusion maps, to discover conserved quantities that vary across trajectories sampled from a dynamical system. We test this new approach on a variety of physical systems—including conservative Hamiltonian systems, dissipative systems, and spatiotemporal systems—and demonstrate that our manifold learning method is able to both identify the number of conserved quantities and extract their values. Using tools from optimal transport theory and manifold learning, our proposed method provides a direct geometric approach to identifying conservation laws that is both robust and interpretable without requiring an explicit model of the system nor accurate time information.

Inferring subhalo effective density slopes from strong lensing observations with neural likelihood-ratio estimation
Gemma Zhang, Siddharth Mishra-Sharma, Cora Dvorkin
Monthly Notices of the Royal Astronomical Society, 2022, Volume 517, Issue 3 [ arXiv:2208.13796 | code ] published
Astrophysics

Abstract Strong gravitational lensing has emerged as a promising approach for probing dark matter models on sub-galactic scales. Recent work has proposed the subhalo effective density slope as a more reliable observable than the commonly used subhalo mass function. The subhalo effective density slope is a measurement independent of assumptions about the underlying density profile and can be inferred for individual subhalos through traditional sampling methods. To go beyond individual subhalo measurements, we leverage recent advances in machine learning and introduce a neural likelihood-ratio estimator to infer an effective density slope for populations of subhalos. We demonstrate that our method is capable of harnessing the statistical power of multiple subhalos (within and across multiple images) to distinguish between characteristics of different subhalo populations. The computational efficiency warranted by the neural likelihood-ratio estimator over traditional sampling enables statistical studies of dark matter perturbers and is particularly useful as we expect an influx of strong lensing systems from upcoming surveys.

Uncovering dark matter density profiles in dwarf galaxies with graph neural networks
Tri Nguyễn, Siddharth Mishra-Sharma, Reuel Williams, Lina Necib
Physical Review D, 202, Volume 107, Issue 4 [ arXiv:2208.12825 | code ] published
Astrophysics

Abstract Dwarf galaxies are small, dark matter-dominated galaxies, some of which are embedded within the Milky Way. Their lack of baryonic matter (e.g., stars and gas) makes them perfect test beds for probing the properties of dark matter -- understanding the spatial dark matter distribution in these systems can be used to constrain microphysical dark matter interactions that influence the formation and evolution of structures in our Universe. We introduce a new method that leverages simulation-based inference and graph-based machine learning in order to infer the dark matter density profiles of dwarf galaxies from observable kinematics of stars gravitationally bound to these systems. Our approach aims to address some of the limitations of established methods based on dynamical Jeans modeling. We show that this novel method can place stronger constraints on dark matter profiles and, consequently, has the potential to weigh in on some of the ongoing puzzles associated with the small-scale structure of dark matter halos, such as the core-cusp discrepancy.

Sampling QCD field configurations with gauge-equivariant flow models
Ryan Abbott, Michael S. Albergo, Aleksandar Botev, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Gurtej Kanwar, Alexander G. D. G. Matthews, Sébastien Racanière, Ali Razavi, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
[ arXiv:2208.03832 ]
Theoretical Physics

Abstract Machine learning methods based on normalizing flows have been shown to address important challenges, such as critical slowing-down and topological freezing, in the sampling of gauge field configurations in simple lattice field theories. A critical question is whether this success will translate to studies of QCD. This Proceedings presents a status update on advances in this area. In particular, it is illustrated how recently developed algorithmic components may be combined to construct flow-based sampling algorithms for QCD in four dimensions. The prospects and challenges for future use of this approach in at-scale applications are summarized.

Neural Embedding: Learning the Embedding of the Manifold of Physics Data
Sang Eon Park, Philip Harris, Bryan Ostdiek
Journal of High Energy Physics, 2023, Volume 2023, Article 108 [ arXiv:2208.05484 ] published
Experimental Physics

Abstract In this paper, we present a method of embedding physics data manifolds with metric structure into lower dimensional spaces with simpler metrics, such as Euclidean and Hyperbolic spaces. We then demonstrate that it can be a powerful step in the data analysis pipeline for many applications. Using progressively more realistic simulated collisions at the Large Hadron Collider, we show that this embedding approach learns the underlying latent structure. With the notion of volume in Euclidean spaces, we provide for the first time a viable solution to quantifying the true search capability of model agnostic search algorithms in collider physics (i.e. anomaly detection). Finally, we discuss how the ideas presented in this paper can be employed to solve many practical challenges that require the extraction of physically meaningful representations from information in complex high dimensional datasets.

Confinement in non-Abelian lattice gauge theory via persistent homology
Daniel Spitz, Julian M. Urban, Jan M. Pawlowski
Physics Review D 2023, Volume 107, Issue 3 [ arXiv:2208.03955 ] published
Theoretical Physics

Abstract We investigate the structure of confining and deconfining phases in SU(2) lattice gauge theory via persistent homology, which gives us access to the topology of a hierarchy of combinatorial objects constructed from given data. Specifically, we use filtrations by traced Polyakov loops, topological densities, holonomy Lie algebra fields, as well as electric and magnetic fields. This allows for a comprehensive picture of confinement. In particular, topological densities form spatial lumps which show signatures of the classical probability distribution of instanton-dyons. Signatures of well-separated dyons located at random positions are encoded in holonomy Lie algebra fields, following the semi-classical temperature dependence of the instanton appearance probability. Debye screening discriminating between electric and magnetic fields is visible in persistent homology and pronounced at large gauge coupling. All employed constructions are gauge-invariant without a priori assumptions on the configurations under study. This work showcases the versatility of persistent homology for statistical and quantum physics studies, barely explored to date.

Robust Clustering of the Local Milky Way Stellar Kinematic Substructures with Gaia eDR3
Xiaowei Ou, Lina Necib, Anna Frebel
Royal Astronomical Society, 2023, Volume 521, Issue 2 [ arXiv:2208.01056 ] published
Astrophysics

Abstract We apply the clustering algorithm HDBSCAN on the Gaia early third data release astrometry combined with the Gaia second data release radial velocity measurements of almost 5.5 million stars to identify the local stellar kinematic substructures in the solar neighborhood. Understanding these structures helps build a more complete picture of the formation of the Milky Way, as well as an empirical phase space distribution of dark matter that would inform detection experiments. The main goal of this study is to provide a list of the most stable clusters, by taking into account the measurement uncertainties and studying the stability of the clustering results. We apply the clustering algorithm in two spaces, in velocity space in order to study recently accreted structures, and in action-angle space to find phase-mixed structures. We find 23 (6) robust clusters in velocity space (action-angle space) that are consistently not associated with noise. They are attributed to the known structures: the Gaia Sausage-Enceladus, the Helmi Stream, and globular cluster NGC 3201 are found in both spaces, while NGC 104 and the thick disk (Sequoia) are identified in velocity space (action-angle space). We discuss the kinematic properties of these structures and study whether many of the small clusters belong to a similar larger cluster based on their chemical abundances. Although we do not identify any new structures, we find that the HDBSCAN member selection of already known structures is unstable to input kinematics of the stars when resampled within their uncertainties. We therefore present the most stable subset of local kinematic structures, which are consistently identified by the clustering algorithm, and emphasize the need to take into account error propagation during both the manual and automated identification of stellar structures, both for existing ones as well as future discoveries. (abridged)

Characterizing the Expected Behavior of Non-Poissonian Template Fitting
Luis Gabriel C. Bariuan, Tracy R. Slatyer
Physical Review D, 2023, Volume 107, Issue 10–15 [ arXiv:2207.13097 ] published
Astrophysics

Abstract We have performed a systematic study of the statistical behavior of non-Poissonian template fitting (NPTF), a method designed to analyze and characterize unresolved point sources in general counts datasets. In this paper, we focus on the properties and characteristics of the Fermi-LAT gamma-ray data set. In particular, we have simulated and analyzed gamma-ray sky maps under varying conditions of exposure, angular resolution, pixel size, energy window, event selection, and source brightness. We describe how these conditions affect the sensitivity of NPTF to the presence of point sources, for inner-galaxy studies of point sources within the Galactic Center excess, and for the simplified case of isotropic emission. We do not find opportunities for major gains in sensitivity from varying these choices, within the range available with current Fermi-LAT data. We provide an analytic estimate of the NPTF sensitivity to point sources for the case of isotropic emission and perfect angular resolution, and find good agreement with our numerical results for that case.

Reconstructing Cosmological Initial Conditions from Late-Time Structure with Convolutional Neural Networks
Christopher J. Shallue, Daniel J. Eisenstein
Monthly Notices of the Royal Astronomical Society, 2023, Volume 520, Issue 4 [ arXiv:2207.12511 | code ] published
Astrophysics

Abstract We present a method to reconstruct the initial linear-regime matter density field from the late-time non-linearly evolved density field in which we channel the output of standard first-order reconstruction to a convolutional neural network (CNN). Our method shows dramatic improvement over the reconstruction of either component alone. We show why CNNs are not well-suited for reconstructing the initial density directly from the late-time density: CNNs are local models, but the relationship between initial and late-time density is not local. Our method leverages standard reconstruction as a preprocessing step, which inverts bulk gravitational flows sourced over very large scales, transforming the residual reconstruction problem from long-range to local and making it ideally suited for a CNN. We develop additional techniques to account for redshift distortions, which warp the density fields measured by galaxy surveys. Our method improves the range of scales of high-fidelity reconstruction by a factor of 2 in wavenumber above standard reconstruction, corresponding to a factor of 8 increase in the number of well-reconstructed modes. In addition, our method almost completely eliminates the anisotropy caused by redshift distortions. As galaxy surveys continue to map the Universe in increasingly greater detail, our results demonstrate the opportunity offered by CNNs to untangle the non-linear clustering at intermediate scales more accurately than ever before.

Bounding generalization error with input compression: An empirical study with infinite-width networks
Angus Galloway, Anna Golubeva, Mahmoud Salem, Mihai Nica, Yani Ioannou, Graham W. Taylor
[ arXiv:2207.09408 ]
Foundational AI

Abstract Estimating the Generalization Error (GE) of Deep Neural Networks (DNNs) is an important task that often relies on availability of held-out data. The ability to better predict GE based on a single training set may yield overarching DNN design principles to reduce a reliance on trial-and-error, along with other performance assessment advantages. In search of a quantity relevant to GE, we investigate the Mutual Information (MI) between the input and final layer representations, using the infinite-width DNN limit to bound MI. An existing input compression-based GE bound is used to link MI and GE. To the best of our knowledge, this represents the first empirical study of this bound. In our attempt to empirically falsify the theoretical bound, we find that it is often tight for best-performing models. Furthermore, it detects randomization of training labels in many cases, reflects test-time perturbation robustness, and works well given only few training samples. These results are promising given that input compression is broadly applicable where MI can be estimated with confidence.

Modeling early-universe energy injection with Dense Neural Networks
Yitian Sun, Tracy R. Slatyer
Physical Review D, Volume 107, Article 063541 [ arXiv:2207.06425 | code ] published
Astrophysics

Abstract We show that Dense Neural Networks can be used to accurately model the cooling of high-energy particles in the early universe, in the context of the public code package DarkHistory. DarkHistory self-consistently computes the temperature and ionization history of the early universe in the presence of exotic energy injections, such as might arise from the annihilation or decay of dark matter. The original version of DarkHistory uses large pre-computed transfer function tables to evolve photon and electron spectra in redshift steps, which require a significant amount of memory and storage space. We present a light version of DarkHistory that makes use of simple Dense Neural Networks to store and interpolate the transfer functions, which performs well on small computers without heavy memory or storage usage. This method anticipates future expansion with additional parametric dependence in the transfer functions without requiring exponentially larger data tables..

Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions
Ryan Abbott, Michael S. Albergo, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Betsy Tian, Julian M. Urban
Physical Review D, 2022, Volume 106, Issue 7 [ arXiv:2207.08945 ] published
Theoretical Physics

Abstract This work presents gauge-equivariant architectures for flow-based sampling in fermionic lattice field theories using pseudofermions as stochastic estimators for the fermionic determinant. This is the default approach in state-of-the-art lattice field theory calculations, making this development critical to the practical application of flow models to theories such as QCD. Methods by which flow-based sampling approaches can be improved via standard techniques such as even/odd preconditioning and the Hasenbusch factorization are also outlined. Numerical demonstrations in two-dimensional U(1) and SU(3) gauge theories with Nf=2 flavors of fermions are provided.

Radio excess from stimulated dark matter decay
Andrea Caputo,Hongwan Liu, Siddharth Mishra-Sharma, Maxim Pospelov, Joshua T. Ruderman
Physical Review D, 2023, Volume 107, Issue 12, 2023 [ arXiv:2206.07713 ] published
Astrophysics

Abstract Despite an intense theoretical and experimental effort over the past decade, observations of the extragalactic radio background at multiple frequencies below 10 GHz are not understood in terms of known radio sources and may represent a sign of new physics. In this paper, we identify a new class of dark sector models with feebly interacting particles, where dark photons oscillate into ordinary photons that contribute to the radio background. Our scenario can explain both the magnitude and the spectral index of the radio background, while being consistent with other cosmological and astrophysical constraints. These models predict new relativistic degrees of freedom and spectral distortions of the cosmic microwave background, which could be detected in the next generation of experiments.

Deep Learning and Symbolic Regression for Discovering Parametric Equations
Michael Zhang, Samuel Kim, Peter Y. Lu, Marin Soljačić
IEEE Transactions on Neural Networks and Learning Systems, vol. 35, no. 11, pp. 16775-16787, Nov. 2024 [ arXiv:2207.00529 | code ] published
Foundational AI

Abstract Symbolic regression is a machine learning technique that can learn the governing formulas of data and thus has the potential to transform scientific discovery. However, symbolic regression is still limited in the complexity and dimensionality of the systems that it can analyze. Deep learning on the other hand has transformed machine learning in its ability to analyze extremely complex and high-dimensional datasets. We propose a neural network architecture to extend symbolic regression to parametric systems where some coefficient may vary but the structure of the underlying governing equation remains constant. We demonstrate our method on various analytic expressions, ODEs, and PDEs with varying coefficients and show that it extrapolates well outside of the training domain. The neural network-based architecture can also integrate with other deep learning architectures so that it can analyze high-dimensional data while being trained end-to-end. To this end we integrate our architecture with convolutional neural networks to analyze 1D images of varying spring systems.

Strong Lensing Source Reconstruction Using Continuous Neural Fields
Siddharth Mishra-Sharma, Ge Yang
[ arXiv:2206.14820 | code ] published
Astrophysics

Abstract From the nature of dark matter to the rate of expansion of our Universe, observations of distant galaxies distorted through strong gravitational lensing have the potential to answer some of the major open questions in astrophysics. Modeling galaxy-galaxy strong lensing observations presents a number of challenges as the exact configuration of both the background source and foreground lens galaxy is unknown. A timely call, prompted by a number of upcoming surveys anticipating high-resolution lensing images, demands methods that can efficiently model lenses at their full complexity. In this work, we introduce a method that uses continuous neural fields to non-parametrically reconstruct the complex morphology of a source galaxy while simultaneously inferring a distribution over foreground lens galaxy configurations. We demonstrate the efficacy of our method through experiments on simulated data targeting high-resolution lensing images similar to those anticipated in near-future astrophysical surveys.

The Dark Energy Camera Plane Survey 2 (DECaPS2): More Sky, Less Bias, and Better Uncertainties
A. K. Saydjari, E. F. Schlafly, D. Lang, A. M. Meisner, G. M. Green, C. Zucker, I. Zelko, J. S. Speagle, T. Daylan, A. Lee, F. Valdes, D. Schlegel, D. P. Finkbeiner
The Astrophysical Journal Supplement Series, 2023, Vol 264, Number 2 [ arXiv:2206.11909 | code ] published
Astrophysics

Abstract Deep optical and near-infrared imaging of the entire Galactic plane is essential for understanding our Galaxy's stars, gas, and dust. The second data release of the DECam Plane Survey (DECaPS2) extends the five-band optical and near-infrared survey of the southern Galactic plane to cover 6.5% of the sky, |b| < 10° and 6° > l > -124°, complementary to coverage by Pan-STARRS1. Typical single-exposure effective depths, including crowding effects and other complications, are 23.5, 22.6, 22.1, 21.6, and 20.8 mag in g, r, i, z, and Y bands, respectively, with around 1 arcsecond seeing. The survey comprises 3.32 billion objects built from 34 billion detections in 21.4 thousand exposures, totaling 260 hours open shutter time on the Dark Energy Camera (DECam) at Cerro Tololo. The data reduction pipeline features several improvements, including the addition of synthetic source injection tests to validate photometric solutions across the entire survey footprint. A convenient functional form for the detection bias in the faint limit was derived and leveraged to characterize the photometric pipeline performance. A new post-processing technique was applied to every detection to de-bias and improve uncertainty estimates of the flux in the presence of structured backgrounds, specifically targeting nebulosity. The images and source catalogs are publicly available at this http URL: http://decaps.skymaps.info/

Simplifying Polylogarithms with Machine Learning
Aurélien Dersy, Matthew D. Schwartz, Xiaoyuan Zhang
International Journal of Data Science in the Mathematical Sciences, Vol. 01, No. 02, pp. 135-179 (2023) [ arXiv:2206.04115 ] published
Theoretical Physics

Abstract Polylogrithmic functions, such as the logarithm or dilogarithm, satisfy a number of algebraic identities. For the logarithm, all the identities follow from the product rule. For the dilogarithm and higher-weight classical polylogarithms, the identities can involve five functions or more. In many calculations relevant to particle physics, complicated combinations of polylogarithms often arise from Feynman integrals. Although the initial expressions resulting from the integration usually simplify, it is often difficult to know which identities to apply and in what order. To address this bottleneck, we explore to what extent machine learning methods can help. We consider both a reinforcement learning approach, where the identities are analogous to moves in a game, and a transformer network approach, where the problem is viewed analogously to a language-translation task. While both methods are effective, the transformer network appears more powerful and holds promise for practical use in symbolic manipulation tasks in mathematical physics.

Toward a more accurate 3D atlas of C. elegans neurons
Michael Skuhersky, Tailin Wu, Eviatar Yemini, Amin Nejatbakhsh, Edward Boyden & Max Tegmark
BMC Bioinformatics, Volume 23, Article 195 [ ] published
Foundational AI

Abstract Determining cell identity in volumetric images of tagged neuronal nuclei is an ongoing challenge in contemporary neuroscience. Frequently, cell identity is determined by aligning and matching tags to an “atlas” of labeled neuronal positions and other identifying characteristics. Previous analyses of such C. elegans datasets have been hampered by the limited accuracy of such atlases, especially for neurons present in the ventral nerve cord, and also by time-consuming manual elements of the alignment process.

Stable Object Reorientation using Contact Plane Registration
Richard Li, Carlos Esteves, Ameesh Makadia, Pulkit Agrawal
International Conference on Robotics and Automation, 2022 [ arXiv:2208.08962 | code ] published
Foundational AI

Abstract We present a system for accurately predicting stable orientations for diverse rigid objects. We propose to overcome the critical issue of modelling multimodality in the space of rotations by using a conditional generative model to accurately classify contact surfaces. Our system is capable of operating from noisy and partially-observed pointcloud observations captured by real world depth cameras. Our method substantially outperforms the current state-of-the-art systems on a simulated stacking task requiring highly accurate rotations, and demonstrates strong sim2real zero-shot transfer results across a variety of unseen objects on a real world reorientation task.

Revealing the Milky Way’s Most Recent Major Merger with a Gaia EDR3 Catalog of Machine-Learned Line-of-Sight Velocities
Adriana Dropulic, Hongwan Liu, Bryan Ostdiek, Mariangela Lisanti
Monthly Notices of the Royal Astronomical Society, May 2023, Volume 521, Issue 2 [ arXiv:2205.12278 ] published
Astrophysics

Abstract Machine learning can play a powerful role in inferring missing line-of-sight velocities from astrometry in surveys such as Gaia. In this paper, we apply a neural network to Gaia Early Data Release 3 (EDR3) and obtain line-of-sight velocities and associated uncertainties for ~92 million stars. The network, which takes as input a star's parallax, angular coordinates, and proper motions, is trained and validated on ~6.4 million stars in Gaia with complete phase-space information. The network's uncertainty on its velocity prediction is a key aspect of its design; by properly convolving these uncertainties with the inferred velocities, we obtain accurate stellar kinematic distributions. As a first science application, we use the new network-completed catalog to identify candidate stars that belong to the Milky Way's most recent major merger, Gaia-Sausage-Enceladus (GSE). We present the kinematic, energy, angular momentum, and spatial distributions of the ~450,000 GSE candidates in this sample, and also study the chemical abundances of those with cross matches to GALAH and APOGEE. The network's predictive power will only continue to improve with future Gaia data releases as the training set of stars with complete phase-space information grows. This work provides a first demonstration of how to use machine learning to exploit high-dimensional correlations on data to infer line-of-sight velocities, and offers a template for how to train, validate and apply such a neural network when complete observational data is not available.

Towards Understanding Grokking: An Effective Theory of Representation Learning
Ziming Liu, Ouail Kitouni, Niklas Nolte, Eric J. Michaud, Max Tegmark, Mike Williams
Conference on Neural Information Processing Systems, 2022 [ arXiv:2205.10343 | code ] published
Foundational AI

Abstract We aim to understand grokking, a phenomenon where models generalize long after overfitting their training set. We present both a microscopic analysis anchored by an effective theory and a macroscopic analysis of phase diagrams describing learning performance across hyperparameters. We find that generalization originates from structured representations whose training dynamics and dependence on training set size can be predicted by our effective theory in a toy setting. We observe empirically the presence of four learning phases: comprehension, grokking, memorization, and confusion. We find representation learning to occur only in a 'Goldilocks zone' (including comprehension and grokking) between memorization and confusion. Compared to the comprehension phase, the grokking phase stays closer to the memorization phase, leading to delayed generalization. The Goldilocks phase is reminiscent of 'intelligence from starvation' in Darwinian evolution, where resource limitations drive discovery of more efficient solutions. This study not only provides intuitive explanations of the origin of grokking, but also highlights the usefulness of physics-inspired tools, e.g., effective theories and phase diagrams, for understanding deep learning.

Power Counting Energy Flow Polynomials
Pedro Cal, Jesse Thaler, Wouter J. Waalewijn
Journal of High Energy Physics, 2022, Article 21 [ arXiv:2205.06818 ] published
Theoretical Physics

Abstract Power counting is a systematic strategy for organizing collider observables and their associated theoretical calculations. In this paper, we use power counting to characterize a class of jet substructure observables called energy flow polynomials (EFPs). EFPs provide an overcomplete linear basis for infrared-and-collinear safe jet observables, but it is known that in practice, a small subset of EFPs is often sufficient for specific jet analysis tasks. By applying power counting arguments, we obtain linear relationships between EFPs that hold for quark and gluon jets to a specific order in the power counting. We test these relations in the parton shower generator Pythia, finding excellent agreement. Power counting allows us to truncate the basis of EFPs without affecting performance, which we corroborate through a study of quark-gluon tagging and regression.

Bias and Priors in Machine Learning Calibrations for High Energy Physics
Rikab Gambhir, Benjamin Nachman, Jesse Thaler
Physical Review D, Volume 106, Article 036011 [ arXiv:2205.05084 | code ] published
Experimental Physics

Abstract Machine learning offers an exciting opportunity to improve the calibration of nearly all reconstructed objects in high-energy physics detectors. However, machine learning approaches often depend on the spectra of examples used during training, an issue known as prior dependence. This is an undesirable property of a calibration, which needs to be applicable in a variety of environments. The purpose of this paper is to explicitly highlight the prior dependence of some machine learning-based calibration strategies. We demonstrate how some recent proposals for both simulation-based and data-based calibrations inherit properties of the sample used for training, which can result in biases for downstream analyses. In the case of simulation-based calibration, we argue that our recently proposed Gaussian Ansatz approach can avoid some of the pitfalls of prior dependence, whereas prior-independent data-based calibration remains an open problem.

Disentangling Quarks and Gluons with CMS Open Data
Patrick T. Komiske, Serhii Kryhin, Jesse Thaler
Physical Review D, 2022, Volume 106 Article 094021 [ arXiv:2205.04459 ] published
Theoretical Physics

Abstract We study quark and gluon jets separately using public collider data from the CMS experiment. Our analysis is based on 2.3/fb of proton-proton collisions at 7 TeV, collected at the Large Hadron Collider in 2011. We define two non-overlapping samples via a pseudorapidity cut -- central jets with |eta| < 0.65 and forward jets with |eta| > 0.65 -- and employ jet topic modeling to extract individual distributions for the maximally separable categories. Under certain assumptions, such as sample independence and mutual irreducibility, these categories correspond to "quark" and "gluon" jets, as given by a recently proposed operational definition. We consider a number of different methods for extracting reducibility factors from the central and forward datasets, from which the fractions of quark jets in each sample can be determined. The greatest stability and robustness to statistical uncertainties is achieved by a novel method based on parametrizing the endpoints of a receiver operating characteristic (ROC) curve. To mitigate detector effects, which would otherwise induce unphysical differences between central and forward jets, we use the OmniFold method to perform central value unfolding. As a demonstration of the power of this method, we extract the intrinsic dimensionality of the quark and gluon jet samples, which exhibit Casimir scaling, as expected from the strongly-ordered limit. To our knowledge, this work is the first application of full phase space unfolding to real collider data, and one of the first applications of topic modeling to extract separate quark and gluon distributions at the LHC.

Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics
Rikab Gambhir, Benjamin Nachman, Jesse Thaler
Physical Review Letters, 2022, Volume 129, Article 082001 [ arXiv:2205.03413 | code ] published
Experimental Physics

Abstract Calibration is a common experimental physics problem, whose goal is to infer the value and uncertainty of an unobservable quantity Z given a measured quantity X. Additionally, one would like to quantify the extent to which X and Z are correlated. In this paper, we present a machine learning framework for performing frequentist maximum likelihood inference with Gaussian uncertainty estimation, which also quantifies the mutual information between the unobservable and measured quantities. This framework uses the Donsker-Varadhan representation of the Kullback-Leibler divergence -- parametrized with a novel Gaussian Ansatz -- to enable a simultaneous extraction of the maximum likelihood values, uncertainties, and mutual information in a single training. We demonstrate our framework by extracting jet energy corrections and resolution factors from a simulation of the CMS detector at the Large Hadron Collider. By leveraging the high-dimensional feature space inside jets, we improve upon the nominal CMS jet resolution by upwards of 15%.

Rapid Locomotion via Reinforcement Learning
Gabriel B. Margolis, Ge Yang, Kartik Paigwar, Tao Chen, Pulkit Agrawal
[ arXiv:2205.02824 ]
Foundational AI

Abstract Agile maneuvers such as sprinting and high-speed turning in the wild are challenging for legged robots. We present an end-to-end learned controller that achieves record agility for the MIT Mini Cheetah, sustaining speeds up to 3.9m/s. This system runs and turns fast on natural terrains like grass, ice, and gravel and responds robustly to disturbances. Our controller is a neural network trained in simulation via reinforcement learning and transferred to the real world. The two key components are (i) an adaptive curriculum on velocity commands and (ii) an online system identification strategy for sim-to-real transfer leveraged from prior work. Videos of the robot’s behaviors are available at https://agility.csail.mit.edu/.

Infinite Variance in Monte Carlo Sampling of Lattice Field Theories
Cagin Yunus, William Detmold
Physical Review D, Volume 106, Article 094506 [ arXiv:2205.01001 ] published
Theoretical Physics

Abstract In Monte Carlo calculations of expectation values in lattice quantum field theories, the stochastic variance of the sampling procedure that is used defines the precision of the calculation for a fixed number of samples. If the variance of an estimator of a particular quantity is formally infinite, or in practice very large compared to the square of the mean, then that quantity can not be reliably estimated using the given sampling procedure. There are multiple scenarios in which this occurs, including in Lattice Quantum Chromodynamics, and a particularly simple example is given by the Gross-Neveu model where Monte Carlo calculations involve the introduction of auxiliary bosonic variables through a Hubbard-Stratonovich (HS) transformation. Here, it is shown that the variances of HS estimators for classes of operators involving fermion fields are divergent in this model and an even simpler zero-dimensional analogue. To correctly estimate these observables, two alternative sampling methods are proposed and numerically investigated.

Going Beyond the Galaxy Power Spectrum: an Analysis of BOSS Data with Wavelet Scattering Transforms
Georgios Valogiannis, Cora Dvorkin
Physical Review D, 2022, Volume 106, Article 103509 [ arXiv:2204.13717 ] published
Astrophysics

Abstract We perform the first application of the wavelet scattering transform (WST) on actual galaxy observations, through a WST analysis of the BOSS DR12 CMASS dataset. We lay out the detailed procedure on how to capture all necessary layers of realism for an application on data obtained from a spectroscopic survey, including the effects of redshift-space anisotropy, non-trivial survey geometry, the shortcomings of the dataset through a set of systematic weights and the Alcock-Paczynski distortion effect. In order to capture the cosmological dependence of the WST, we use galaxy mocks obtained from the state-of-the-art ABACUSSUMMIT simulations, tuned to match the anisotropic correlation function of the BOSS CMASS sample in the redshift range 0.46<z<0.60. Using our theory model for the WST coefficients, as well as for the first 2 multipoles of the galaxy power spectrum, that we use as reference, we perform a likelihood analysis of the CMASS data and obtain the posterior probability distributions of 4 cosmological parameters, {ωb,ωc,ns,σ8}, as well as the Hubble constant, derived from a fixed value of the angular size of the sound horizon at last scattering measured by the Planck satellite, all of which are marginalized over the 7 nuisance parameters of the Halo Occupation Distribution model. The WST is found to deliver a substantial improvement in the values of the predicted 1σ errors compared to the regular power spectrum, which are tighter by a factor in the range 3−6 in the case of flat and uninformative priors and by a factor of 4−28, when a Big Bang Nucleosynthesis prior is applied on the value of ωb. Furthermore, in the latter case, we obtain a 0.6% measurement of the Hubble constant. Our results are investigative and subject to certain approximations in our analysis, that we discuss in the text.

DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
Yung-Sung Chuang, Rumen Dangovski, Hongyin Luo, Yang Zhang, Shiyu Chang, Marin Soljačić, Shang-Wen Li, Wen-tau Yin, Yoon Kim, James Glass
[ arXiv:2204.10298 | code ]
Foundational AI

Abstract We propose DiffCSE, an unsupervised contrastive learning framework for learning sentence embeddings. DiffCSE learns sentence embeddings that are sensitive to the difference between the original sentence and an edited sentence, where the edited sentence is obtained by stochastically masking out the original sentence and then sampling from a masked language model. We show that DiffSCE is an instance of equivariant contrastive learning (Dangovski et al., 2021), which generalizes contrastive learning and learns representations that are insensitive to certain types of augmentations and sensitive to other "harmful" types of augmentations. Our experiments show that DiffCSE achieves state-of-the-art results among unsupervised sentence representation learning methods, outperforming unsupervised SimCSE by 2.3 absolute points on semantic textual similarity tasks.

Photometrically-Classified Superluminous Supernovae from the Pan-STARRS1 Medium Deep Survey: A Case Study for Science with Machine Learning-Based Classification
Brian Hsu, Griffin Hosseinzadeh, V. Ashley Villar, Edo Berger
The Astrophysical Journal, 2022, Volume 937, Number 1 [ arXiv:2204.09809 ] published
Astrophysics

Abstract With the upcoming Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST), it is expected that only ∼0.1% of all transients will be classified spectroscopically. To conduct studies of rare transients, such as Type I superluminous supernovae (SLSNe), we must instead rely on photometric classification. In this vein, here we carry out a pilot study of SLSNe from the Pan-STARRS1 Medium-Deep Survey (PS1-MDS) classified photometrically with our SuperRAENN and Superphot algorithms. We first construct a sub-sample of the photometric sample using a list of simple selection metrics designed to minimize contamination and ensure sufficient data quality for modeling. We then fit the multi-band light curves with a magnetar spin-down model using the Modular Open-Source Fitter for Transients (MOSFiT). Comparing the magnetar engine and ejecta parameter distributions of the photometric sample to those of the PS1-MDS spectroscopic sample and a larger literature spectroscopic sample, we find that these samples are overall consistent, but that the photometric sample extends to slower spins and lower ejecta masses, which correspond to lower luminosity events, as expected for photometric selection. While our PS1-MDS photometric sample is still smaller than the overall SLSN spectroscopic sample, our methodology paves the way to an orders-of-magnitude increase in the SLSN sample in the LSST era through photometric selection and study.

Luminous Supernovae: Unveiling a Population Between Superluminous and Normal Core-collapse Supernovae
Sebastian Gomez, Edo Berger, Matt Nicholl, Peter K. Blanchard, Griffin Hosseinzadeh
The Astrophysical Journal, 2022, Volume 941, Number 2 [ arXiv:2204.08486 ] published
Astrophysics

Abstract Stripped-envelope core-collapse supernovae can be divided into two broad classes: the common Type Ib/c supernovae (SNe Ib/c), powered by the radioactive decay of 56Ni, and the rare superluminous supernovae (SLSNe), most likely powered by the spin-down of a magnetar central engine. Up to now, the intermediate regime between these two populations has remained mostly unexplored. Here, we present a comprehensive study of 40 extit{luminous supernovae} (LSNe), SNe with peak magnitudes of Mr=−19 to −20 mag, bound by SLSNe on the bright end and by SNe Ib/c on the dim end. Spectroscopically, LSNe appear to form a continuum between Type Ic SNe and SLSNe. Given their intermediate nature, we model the light curves of all LSNe using a combined magnetar plus radioactive decay model and find that they are indeed intermediate, not only in terms of their peak luminosity and spectra, but also in their rise times, power sources, and physical parameters. We sub-classify LSNe into distinct groups that are either as fast-evolving as SNe Ib/c or as slow-evolving as SLSNe, and appear to be either radioactively or magnetar powered, respectively. Our findings indicate that LSNe are powered by either an over-abundant production of 56Ni or by weak magnetar engines, and may serve as the missing link between the two populations.

Quantification of high dimensional non-Gaussianities and its implication to Fisher analysis in cosmology
Core Francisco Park, Erwan Allys, Francisco Villaescusa-Navarro, Douglas P. Finkbeiner
The Astrophysical Journal, Volume 946, Number 2 [ arXiv:2204.05435 | code ] published
Astrophysics

Abstract It is well known that the power spectrum is not able to fully characterize the statistical properties of non-Gaussian density fields. Recently, many different statistics have been proposed to extract information from non-Gaussian cosmological fields that perform better than the power spectrum. The Fisher matrix formalism is commonly used to quantify the accuracy with which a given statistic can constrain the value of the cosmological parameters. However, these calculations typically rely on the assumption that the likelihood of the considered statistic follows a multivariate Gaussian distribution. In this work we follow Sellentin & Heavens (2017) and use two different statistical tests to identify non-Gaussianities in different statistics such as the power spectrum, bispectrum, marked power spectrum, and wavelet scatering transform (WST). We remove the non-Gaussian components of the different statistics and perform Fisher matrix calculations with the extit{Gaussianized} statistics using Quijote simulations. We show that constraints on the parameters can change by a factor of ∼2 in some cases. We show with simple examples how statistics that do not follow a multivariate Gaussian distribution can achieve artificially tight bounds on the cosmological parameters when using the Fisher matrix formalism. We think that the non-Gaussian tests used in this work represent a powerful tool to quantify the robustness of Fisher matrix calculations and their underlying assumptions. We release the code used to compute the power spectra, bispectra, and WST that can be run on both CPUs and GPUs.

Pareto-optimal clustering with the primal deterministic information bottleneck
Andrew K. Tan, Max Tegmark, Isaac L. Chuang
Entropy, 2022, 24(6) [ arXiv:2204.02489 | code ] published
Foundational AI

Abstract At the heart of both lossy compression and clustering is a trade-off between the fidelity and size of the learned representation. Our goal is to map out and study the Pareto frontier that quantifies this trade-off. We focus on the Deterministic Information Bottleneck (DIB) formulation of lossy compression, which can be interpreted as a clustering problem. To this end, we introduce the {\it primal} DIB problem, which we show results in a much richer frontier than its previously studied dual counterpart. We present an algorithm for mapping out the Pareto frontier of the primal DIB trade-off that is also applicable to most other two-objective clustering problems. We study general properties of the Pareto frontier, and give both analytic and numerical evidence for logarithmic sparsity of the frontier in general. We provide evidence that our algorithm has polynomial scaling despite the super-exponential search space; and additionally propose a modification to the algorithm that can be used where sampling noise is expected to be significant. Finally, we use our algorithm to map the DIB frontier of three different tasks: compressing the English alphabet, extracting informative color classes from natural images, and compressing a group theory inspired dataset, revealing interesting features of frontier, and demonstrating how the structure of the frontier can be used for model selection with a focus on points previously hidden by the cloak of the convex hull.

AI Poincaré 2.0: Machine Learning Conservation Laws from Differential Equations
Ziming Liu, Varun Madhavan, Max Tegmark
Physical Review E, 2022, Volume 106, Article 045307 [ arXiv:2203.12610 ] published
Foundational AI

Abstract We present a machine learning algorithm that discovers conservation laws from differential equations, both numerically (parametrized as neural networks) and symbolically, ensuring their functional independence (a non-linear generalization of linear independence). Our independence module can be viewed as a nonlinear generalization of singular value decomposition. Our method can readily handle inductive biases for conservation laws. We validate it with examples including the 3-body problem, the KdV equation and nonlinear Schrödinger equation.

Unsupervised Semantic Segmentation by Distilling Feature Correspondences
Mark Hamilton, Zhoutong Zhang, Bharath Hariharan, Noah Snavely, William T. Freeman
ICLR 2022 [ arXiv:2203.08414 | code ] published
Foundational AI

Abstract Unsupervised semantic segmentation aims to discover and localize semantically meaningful categories within image corpora without any form of annotation. To solve this task, algorithms must produce features for every pixel that are both semantically meaningful and compact enough to form distinct clusters. Unlike previous works which achieve this with a single end-to-end framework, we propose to separate feature learning from cluster compactification. Empirically, we show that current unsupervised feature learning frameworks already generate dense features whose correlations are semantically consistent. This observation motivates us to design STEGO (Self-supervised Transformer with Energy-based Graph Optimization), a novel framework that distills unsupervised features into high-quality discrete semantic labels. At the core of STEGO is a novel contrastive loss function that encourages features to form compact clusters while preserving their relationships across the corpora. STEGO yields a significant improvement over the prior state of the art, on both the CocoStuff (+14 mIoU) and Cityscapes (+9 mIoU) semantic segmentation challenges.

Categorical Representation Learning and RG flow operators for algorithmic classifiers
Artan Sheshmani, Yizhuang You, Wenbo Fu, Ahmadreza Azizi
Machine Learning Science and Technology, Volume 4, Number 1, Article 015012 [ arXiv:2203.07975 ] published
Foundational AI

Abstract Following the earlier formalism of the categorical representation learning (arXiv:2103.14770) by the first two authors, we discuss the construction of the RG-flow based categorifier. Borrowing ideas from theory of renormalization group flows (RG) in quantum field theory, holographic duality, and hyperbolic geometry, and mixing them with neural ODE's, we construct a new algorithmic natural language processing (NLP) architecture, called the RG-flow categorifier or for short the RG categorifier, which is capable of data classification and generation in all layers. We apply our algorithmic platform to biomedical data sets and show its performance in the field of sequence-to-function mapping. In particular we apply the RG categorifier to particular genomic sequences of flu viruses and show how our technology is capable of extracting the information from given genomic sequences, find their hidden symmetries and dominant features, classify them and use the trained data to make stochastic prediction of new plausible generated sequences associated with new set of viruses which could avoid the human immune system. The content of the current article is part of the recent US patent application submitted by first two authors (U.S. Patent Application No.: 63/313.504).

Flow-based density of states for complex actions
Jan M. Pawlowski, Julian M. Urban
Physical Review D, 2023, Volume 108, Issue 5 [ arXiv:2203.01243 ] published
Theoretical Physics

Abstract Emerging sampling algorithms based on normalizing flows have the potential to solve ergodicity problems in lattice calculations. Furthermore, it has been noted that flows can be used to compute thermodynamic quantities which are difficult to access with traditional methods. This suggests that they are also applicable to the density-of-states approach to complex action problems. In particular, flow-based sampling may be used to compute the density directly, in contradistinction to the conventional strategy of reconstructing it via measuring and integrating the derivative of its logarithm. By circumventing this procedure, the accumulation of errors from the numerical integration is avoided completely and the overall normalization factor can be determined explicitly. In this proof-of-principle study, we demonstrate our method in the context of two-component scalar field theory where the O(2) symmetry is explicitly broken by an imaginary external field. First, we concentrate on the zero-dimensional case which can be solved exactly. We show that with our method, the Lee-Yang zeroes of the associated partition function can be successfully located. Subsequently, we confirm that the flow-based approach correctly reproduces the density computed with conventional methods in one- and two-dimensional models.

Creating Simple, Interpretable Anomaly Detectors for New Physics in Jet Substructure
Layne Bradshaw, Spencer Chang, Bryan Ostdiek
Physical Review D, 2022, Volume 106, Article 035014 [ arXiv:2203.01343 ] published
Theoretical Physics

Abstract Anomaly detection with convolutional autoencoders is a popular method to search for new physics in a model-agnostic manner. These techniques are powerful, but they are still a "black box," since we do not know what high-level physical observables determine how anomalous an event is. To address this, we adapt a recently proposed technique by Faucett this http URL, which maps out the physical observables learned by a neural network classifier, to the case of anomaly detection. We propose two different strategies that use a small number of high-level observables to mimic the decisions made by the autoencoder on background events. Despite the underlying differences in their approach, we find that both strategies have similar ordering performance as the autoencoder and independently use the same five high-level observables. From there, we compare the performance of these networks as anomaly detectors. We find that both strategies perform similarly to the autoencoder across a variety of signals, giving a nontrivial demonstration that learning to order background events transfers to ordering a variety of signal events.

Biological error correction codes generate fault-tolerant neural networks
Alexander Zlokapa, Andrew K. Tan, John M. Martyn, Max Tegmark, Isaac L. Chuang
Physical Review E, 2024, Volume 110, Issue 5 [ arXiv:2202.12887 ] published
Foundational AI

Abstract It has been an open question in deep learning if fault-tolerant computation is possible: can arbitrarily reliable computation be achieved using only unreliable neurons? In the mammalian cortex, analog error correction codes known as grid codes have been observed to protect states against neural spiking noise, but their role in information processing is unclear. Here, we use these biological codes to show that a universal fault-tolerant neural network can be achieved if the faultiness of each neuron lies below a sharp threshold, which we find coincides in order of magnitude with noise observed in biological neurons. The discovery of a sharp phase transition from faulty to fault-tolerant neural computation opens a path towards understanding noisy analog systems in artificial intelligence and neuroscience.

Flow-based sampling in the lattice Schwinger model at criticality
Michael S. Albergo, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Fernando Romero-López, Phiala E. Shanahan, Julian M. Urban
Physical Review D, 2022, Volume 106, Article 014514 [ arXiv:2202.11712 ] published
Theoretical Physics

Abstract Recent results suggest that flow-based algorithms may provide efficient sampling of field distributions for lattice field theory applications, such as studies of quantum chromodynamics and the Schwinger model. In this work, we provide a numerical demonstration of robust flow-based sampling in the Schwinger model at the critical value of the fermion mass. In contrast, at the same parameters, conventional methods fail to sample all parts of configuration space, leading to severely underestimated uncertainties.

Identifying equivalent Calabi–Yau topologies: A discrete challenge from math and physics for machine learning
Vishnu Jejjala, Washington Taylor, Andrew Turner
[ arXiv:2202.07590 ]
Theoretical Physics

Abstract We review briefly the characteristic topological data of Calabi–Yau threefolds and focus on the question of when two threefolds are equivalent through related topological data. This provides an interesting test case for machine learn- ing methodology in discrete mathematics problems motivated by physics.

Topogivity: A Machine-Learned Chemical Rule for Discovering Topological Materials
Andrew Ma, Yang Zhang, Thomas Christensen, Hoi Chun Po, Li Jing, Liang Fu, Marin Soljačić
American Chemical Society Publications [ arXiv:2202.05255 | code ] published
Astrophysics

Abstract Topological materials present unconventional electronic properties that make them attractive for both basic science and next-generation technological applications. The majority of currently-known topological materials have been discovered using methods that involve symmetry-based analysis of the quantum wavefunction. Here we use machine learning to develop a simple-to-use heuristic chemical rule that diagnoses with a high accuracy whether a material is topological using only its chemical formula. This heuristic rule is based on a notion that we term topogivity, a machine-learned numerical value for each element that loosely captures its tendency to form topological materials. We next implement a high-throughput strategy for discovering topological materials based on the heuristic topogivity-rule prediction followed by ab initio validation. This way, we discover new topological materials that are not diagnosable using symmetry indicators, including several that may be promising for experimental observation.

Finite-Volume Pionless Effective Field Theory for Few-Nucleon Systems with Differentiable Programming
Xiangkai Sun, William Detmold, Di Luo, Phiala E. Shanahan
[ arXiv:2202.03530 ]
Theoretical Physics

Abstract Finite-volume pionless effective field theory provides an efficient framework for the extrapolation of nuclear spectra and matrix elements calculated at finite volume in lattice QCD to infinite volume, and to nuclei with larger atomic number. In this work, it is demonstrated how this framework may be implemented via a set of correlated Gaussian wavefunctions optimised using differentiable programming and via solution of a generalised eigenvalue problem. This approach is shown to be significantly more efficient than a stochastic implementation of the variational method based on the same form of correlated Gaussian wavefunctions, yielding comparably accurate representations of the ground-state wavefunctions with an order of magnitude fewer terms. The efficiency of representation allows such calculations to be extended to larger systems than in previous work. The method is demonstrated through calculations of the binding energies of nuclei with atomic number A∈{2,3,4} in finite volume, matched to lattice QCD calculations at quark masses corresponding to mπ=806 MeV, and infinite-volume effective field theory calculations of A∈{2,3,4,5,6} systems based on this matching.

Constraining the Time of Gravitational Wave Emission from Core-Collapse Supernovae
Kiranjyot Gill, Griffin Hosseinzadeh, Edo Berger, Michele Zanolin, Marek Szczepanczyk
The Astrophysical Journal, 2022, Volume 931, Number 2 [ arXiv:2201.03609 ] published
Astrophysics

Abstract The advent of sensitive gravitational wave (GW) detectors, coupled with wide-field, high cadence optical time-domain surveys, raises the possibility of the first join GW-electromagnetic (EM) detections of core-collapse supernovae (CCSNe). For targeted searches of Gas from CCSNe optical observation can be used to increase the sensitivity of the search by restricting the relevant time interval, defined here as the GW search window (GSW). The extent of the GSW is a critical factor in determining the achievable false alarm probability (FAP) for a triggered CCSN search. The ability to constrain the GSW from optical observations depends on how early a CCSN is detected, as well as the ability to model the early optical emission. Here we present several approaches to constrain the GSW, ranging in complexity from model-independent analytical fits of the early light curve, model-dependent fits of the rising or entire light curve, and a new data-driven approach using existing well-sampled CCSN light curves from {\it Kepler} and the Transiting Exoplanet Survey Satellite (TESS). We use these approaches to determine the time of core-collapse and its associated uncertainty (i.e., the GSW). We apply our methods to two Type II See that occurred during LIGO/Virgo Observing Run 3: SN\,2019fcn and SN\,2019ejj (both in the same galaxy at d = 15.7 Mac). Our approach shortens the duration of the GSW and improves the robustness of the GSW compared to techniques used in past GW CCSN searches.

Photometry on Structured Backgrounds: Local Pixelwise Infilling by Regression
Andrew K. Saydjari, Douglas P. Finkbeiner
The Astrophysical Journal, 2022, Volume 933, Number 2 [ arXiv:2201.07246 | code ] published
Astrophysics

Abstract Photometric pipelines struggle to estimate both the flux and flux uncertainty for stars in the presence of structured backgrounds such as filaments or clouds. However, it is exactly stars in these complex regions that are critical to understanding star formation and the structure of the interstellar medium. We develop a method, similar to Gaussian process regression, which we term local pixelwise infilling (LPI). Using a local covariance estimate, we predict the background behind each star and the uncertainty on that prediction in order to improve estimates of flux and flux uncertainty. We show the validity of our model on synthetic data and real dust fields. We further demonstrate that the method is stable even in the crowded field limit. While we focus on optical-IR photometry, this method is not restricted to those wavelengths. We apply this technique to the 34 billion detections in the second data release of the Dark Energy Camera Plane Survey (DECaPS2). In addition to removing many >3σ outliers and improving uncertainty estimates by a factor of ∼2−3 on nebulous fields, we also show that our method is well-behaved on uncrowded fields. The entirely post-processing nature of our implementation of LPI photometry allows it to easily improve the flux and flux uncertainty estimates of past as well as future surveys.

Cracking the Quantum Scaling Limit with Machine Learned Electron Densities
Joshua A. Rackers, Lucas Tecot, Mario Geiger, Tess E. Smidt
[ arXiv:2201.03726 | code ]
Foundational AI

Abstract A long-standing goal of science is to accurately solve the Schrödinger equation for large molecular systems. The poor scaling of current quantum chemistry algorithms on classical computers imposes an effective limit of about a few dozen atoms for which we can calculate molecular electronic structure. We present a machine learning (ML) method to break through this scaling limit and make quantum chemistry calculations of very large systems possible. We show that Euclidean Neural Networks can be trained to predict the electron density with high fidelity from limited data. Learning the electron density allows us to train a machine learning model on small systems and make accurate predictions on large ones. We show that this ML electron density model can break through the quantum scaling limit and calculate the electron density of systems of thousands of atoms with quantum accuracy.

Strictification and gluing of Lagrangian distributions on derived schemes with shifted symplectic forms
Dennis Borisov, Ludmil Katzarkov, Artan Sheshmani, Shing-Tung Yau
Advances in Mathematics, 2024, Volume 438 [ arXiv:1908.00651 ] published
Theoretical Physics

Abstract A strictification result is proved for isotropic distributions on derived schemes equipped with negatively shifted homotopically closed 2-forms. It is shown that any derived scheme over ℂ equipped with a −2-shifted symplectic structure, and having a Hausdorff space of classical points, admits a globally defined Lagrangian distribution as a dg ℂ∞-manifold.

Towards Quantum Simulations in Particle Physics and Beyond on Noisy Intermediate-Scale Quantum Devices
Lena Funcke, Tobias Hartung, Karl Jansen, Stefan Kühn, Manuel Schneider, Paolo Stornati, Xiaoyang Wang
Philosophical Transactions of the Royal Society A [ arXiv:2110.03809 ] published
Theoretical Physics

Abstract We review two algorithmic advances that bring us closer to reliable quantum simulations of model systems in high energy physics and beyond on noisy intermediate-scale quantum (NISQ) devices. The first method is the dimensional expressivity analysis of quantum circuits, which allows for constructing minimal but maximally expressive quantum circuits. The second method is an efficient mitigation of readout errors on quantum devices. Both methods can lead to significant improvements in quantum simulations, e.g., when variational quantum eigensolvers are used.

Invariance Through Latent Alignment
Takuma Yoneda, Ge Yang, Matthew R. Walter, Bradly Stadie
[ arXiv:2112.08526 | code ]
Foundational AI

Abstract A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to better cover what the agent might experience at test time. In many cases, however, it is impossible to know test-time distribution-shift a priori, making these schemes infeasible. In this paper, we introduce a general approach, called Invariance Through Latent Alignment (ILA), that improves the test-time performance of a visuomotor control policy in deployment environments with unknown perceptual variations. ILA performs unsupervised adaptation at deployment-time by matching the distribution of latent features on the target domain to the agent's prior experience, without relying on paired data. Although simple, we show that this idea leads to surprising improvements on a variety of challenging adaptation scenarios, including changes in lighting conditions, the content in the scene, and camera poses. We present results on calibrated control benchmarks in simulation -- the distractor control suite -- and a physical robot under a sim-to-real setup.

Impact of Massive Binary Star and Cosmic Evolution on Gravitational Wave Observations II: Double Compact Object Rates and Properties
Floor S. Broekgaarden, Edo Berger, Simon Stevenson, Stephen Justham, Ilya Mandel, Martyna Churślińska, Like A. C. van Son, Tom Wagg, Alejandro Vigna-Gómez, Selma E. De Mink, Debatri Chattopadhyay, Coenraad J. Neijssel
Monthly Notices of the Royal Astronomical Society, 2022, Volume 516, Issue 4, Pages 5737–5761 [ arXiv:2112.05763 | code ] published
Astrophysics

Abstract Making the most of the rapidly increasing population of gravitational-wave detections of black hole (BH) and neutron star (NS) mergers requires comparing observations with population synthesis predictions. In this work we investigate the combined impact from the key uncertainties in population synthesis modelling of the isolated binary evolution channel: the physical processes in massive binary-star evolution and the star formation history as a function of metallicity, Z, and redshift z,S(Z,z). Considering these uncertainties we create 560 different publicly available model realizations and calculate the rate and distribution characteristics of detectable BHBH, BHNS, and NSNS mergers. We find that our stellar evolution and S(Z,z) variations can impact the predicted intrinsic and detectable merger rates by factors 102-104. We find that BHBH rates are dominantly impacted by S(Z,z) variations, NSNS rates by stellar evolution variations and BHNS rates by both. We then consider the combined impact from all uncertainties considered in this work on the detectable mass distribution shapes (chirp mass, individual masses and mass ratio). We find that the BHNS mass distributions are predominantly impacted by massive binary-star evolution changes. For BHBH and NSNS we find that both uncertainties are important. We also find that the shape of the delay time and birth metallicity distributions are typically dominated by the choice of S(Z,z) for BHBH, BHNS and NSNS. We identify several examples of robust features in the mass distributions predicted by all 560 models, such that we expect more than 95% of BHBH detections to contain a BH ≳8M⊙ and have mass ratios ≲4. Our work demonstrates that it is essential to consider a wide range of allowed models to study double compact object merger rates and properties.

SymmetryGAN: Symmetry Discovery with Deep Learning
Krish Desai, Benjamin Nachman, Jesse Thaler
Physical. Rev. D, 2022, 105:096031 [ arXiv:2112.05722 | code ] published
Theoretical Physics

Abstract What are the symmetries of a dataset? Whereas the symmetries of an individual data element can be characterized by its invariance under various transformations, the symmetries of an ensemble of data elements are ambiguous due to Jacobian factors introduced while changing coordinates. In this paper, we provide a rigorous statistical definition of the symmetries of a dataset, which involves inertial reference densities, in analogy to inertial frames in classical mechanics. We then propose SymmetryGAN as a novel and powerful approach to automatically discover symmetries using a deep learning method based on generative adversarial networks (GANs). When applied to Gaussian examples, SymmetryGAN shows excellent empirical performance, in agreement with expectations from the analytic loss landscape. SymmetryGAN is then applied to simulated dijet events from the Large Hadron Collider (LHC) to demonstrate the potential utility of this method in high energy collider physics applications. Going beyond symmetry discovery, we consider procedures to infer the underlying symmetry group from empirical data.

Neural Descriptor Fields: SE(3) Equivariant Object Representations for Manipulation
Anthony Simeonov, Yilun Du, Andrea Tagliasacchi, Joshua B. Tenenbaum, Alberto Rodriguez, Pulkit Agrawal, Vincent Sitzmann
International Conference on Robotics and Automation 2022 [ arXiv:2112.05124 | code ] published
Foundational AI

Abstract We present Neural Descriptor Fields (NDFs), an object representation that encodes both points and relative poses between an object and a target (such as a robot gripper or a rack used for hanging) via category-level descriptors. We employ this representation for object manipulation, where given a task demonstration, we want to repeat the same task on a new object instance from the same category. We propose to achieve this objective by searching (via optimization) for the pose whose descriptor matches that observed in the demonstration. NDFs are conveniently trained in a self-supervised fashion via a 3D auto-encoding task that does not rely on expert-labeled keypoints. Further, NDFs are SE(3)-equivariant, guaranteeing performance that generalizes across all possible 3D object translations and rotations. We demonstrate learning of manipulation tasks from few (5-10) demonstrations both in simulation and on a real robot. Our performance generalizes across both object instances and 6-DoF object poses, and significantly outperforms a recent baseline that relies on 2D descriptors.

PQ Axiverse
Mehmet Demirtas, Naomi Gendler, Cody Long, Liam McAllister, Jakob Moritz
Journal of High Energy Physics 2023, Volume 2023, Article number 92 [ arXiv:2112.04503 ] published
Theoretical Physics

Abstract We show that the strong CP problem is solved in a large class of compactifications of string theory. The Peccei-Quinn mechanism solves the strong CP problem if the CP-breaking effects of the ultraviolet completion of gravity and of QCD are small compared to the CP-preserving axion potential generated by low-energy QCD instantons. We characterize both classes of effects. To understand quantum gravitational effects, we consider an ensemble of flux compactifications of type IIB string theory on orientifolds of Calabi-Yau hypersurfaces in the geometric regime, taking a simple model of QCD on D7-branes. We show that the D-brane instanton contribution to the neutron electric dipole moment falls exponentially in N4, with N the number of axions. In particular, this contribution is negligible in all models in our ensemble with N>17. We interpret this result as a consequence of large N effects in the geometry that create hierarchies in instanton actions and also suppress the ultraviolet cutoff. We also compute the CP breaking due to high-energy instantons in QCD. In the absence of vectorlike pairs, we find contributions to the neutron electric dipole moment that are not excluded, but that could be accessible to future experiments if the scale of supersymmetry breaking is sufficiently low. The existence of vectorlike pairs can lead to a larger dipole moment. Finally, we show that a significant fraction of models are allowed by standard cosmological and astrophysical constraints.

Building Quantum Field Theories Out of Neurons
James Halverson
[ arXiv:2112.04527 ]
Theoretical Physics

Abstract An approach to field theory is studied in which fields are comprised of N constituent random neurons. Gaussian theories arise in the infinite-N limit when neurons are independently distributed, via the Central Limit Theorem, while interactions arise due to finite-N effects or non-independently distributed neurons. Euclidean-invariant ensembles of neurons are engineered, with tunable two-point function, yielding families of Euclidean-invariant field theories. Some Gaussian, Euclidean invariant theories are reflection positive, which allows for analytic continuation to a Lorentz-invariant quantum field theory. Examples are presented that yield dual theories at infinite-N, but have different symmetries at finite-N. Landscapes of classical field configurations are determined by local maxima of parameter distributions. Predictions arise from mixed field-neuron correlators. Near-Gaussianity is exhibited at large-N, potentially explaining a feature of field theories in Nature.

Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields
Dor Verbin, Peter Hedman, Ben Mildenhall, Todd Zickler, Jonathan T. Barron, Pratul P. Srinivasan
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition [ arXiv:2112.03907 ] published
Foundational AI

Abstract Neural Radiance Fields (NeRF) is a popular view synthesis technique that represents a scene as a continuous volumetric function, parameterized by multilayer perceptrons that provide the volume density and view-dependent emitted radiance at each location. While NeRF-based techniques excel at representing fine geometric structures with smoothly varying view-dependent appearance, they often fail to accurately capture and reproduce the appearance of glossy surfaces. We address this limitation by introducing Ref-NeRF, which replaces NeRF's parameterization of view-dependent outgoing radiance with a representation of reflected radiance and structures this function using a collection of spatially-varying scene properties. We show that together with a regularizer on normal vectors, our model significantly improves the realism and accuracy of specular reflections. Furthermore, we show that our model's internal representation of outgoing radiance is interpretable and useful for scene editing.

Machine Learning in Nuclear Physics
Amber Boehnlein, Markus Diefenthaler, Nobuo Sato, Malachi Schram, Veronique Ziegler, Cristiano Fanelli, Morten Hjorth-Jensen, Tanja Horn, Michelle P. Kuchera, Dean Lee, Witold Nazarewicz, Peter Ostroumov, Kostas Orginos, Alan Poon, Xin-Nian Wang, Alexander Scheinker, Michael S. Smith, and Long-Gang Pang
Reviews of Modern Physics, 2022, Volume 94, Article 031003 [ arXiv:2112.02309 ] published
Theoretical Physics

Abstract Advances in machine learning methods provide tools that have broad applicability in scientific research. These techniques are being applied across the diversity of nuclear physics research topics, leading to advances that will facilitate scientific discoveries and societal applications. This Colloquium provides a snapshot of nuclear physics research, which has been transformed by machine learning techniques.

Infinite Neural Network Quantum States
Di Luo, James Halverson
Machine Learning: Science and Technology, 2023, Volume 4, Number 2 [ arXiv:2112.00723 ] published
Theoretical Physics

Abstract We study infinite limits of neural network quantum states (∞-NNQS), which exhibit representation power through ensemble statistics, and also tractable gradient descent dynamics. Ensemble averages of Renyi entropies are expressed in terms of neural network correlators, and architectures that exhibit volume-law entanglement are presented. A general framework is developed for studying the gradient descent dynamics of neural network quantum states (NNQS), using a quantum state neural tangent kernel (QS-NTK). For ∞-NNQS the training dynamics is simplified, since the QS-NTK becomes deterministic and constant. An analytic solution is derived for quantum state supervised learning, which allows an ∞-NNQS to recover any target wavefunction. Numerical experiments on finite and infinite NNQS in the transverse field Ising model and Fermi Hubbard model demonstrate excellent agreement with theory. ∞-NNQS opens up new opportunities for studying entanglement and training dynamics in other physics applications, such as in finding ground states.

Substructure Detection Reanalyzed: Dark Perturber shown to be a Line-of-Sight Halo
Atınç Çağan Şengül, Cora Dvorkin, Bryan Ostdiek, Arthur Tsang
Monthly Notices of the Royal Astronomical Society, 2022, Volume 515, Issue 3, Pages 4391–4401 [ arXiv:2112.00749 | code ] published
Astrophysics

Abstract Observations of structure at sub-galactic scales are crucial for probing the properties of dark matter, which is the dominant source of gravity in the universe. It will become increasingly important for future surveys to distinguish between line-of-sight halos and subhalos to avoid wrong inferences on the nature of dark matter. We reanalyze a sub-galactic structure (in lens JVAS B1938+666) that has been previously found using the gravitational imaging technique in galaxy-galaxy lensing systems. This structure has been assumed to be a satellite in the halo of the main lens galaxy. We fit the redshift of the perturber of the system as a free parameter, using the multi-plane thin-lens approximation, and find that the redshift of the perturber is zint=1.22+0.11−0.11 (with a main lens redshift of z=0.881). Our analysis indicates that this structure is more massive than the previous result by more than an order of magnitude. This constitutes the first dark perturber shown to be a line-of-sight halo with a gravitational lensing method.

Robust and Provably Monotonic Networks
Ouail Kitouni, Niklas Nolte, Mike Williams
Machine Learning: Science and Technology, Volume 4, Number 3, 2023 [ arXiv:2112.00038 | code ] published
Experimental Physics

Abstract The Lipschitz constant of the map between the input and output space represented by a neural network is a natural metric for assessing the robustness of the model. We present a new method to constrain the Lipschitz constant of dense deep learning models that can also be generalized to other architectures. The method relies on a simple weight normalization scheme during training that ensures the Lipschitz constant of every layer is below an upper limit specified by the analyst. A simple residual connection can then be used to make the model monotonic in any subset of its inputs, which is useful in scenarios where domain knowledge dictates such dependence. Examples can be found in algorithmic fairness requirements or, as presented here, in the classification of the decays of subatomic particles produced at the CERN Large Hadron Collider. Our normalization is minimally constraining and allows the underlying architecture to maintain higher expressiveness compared to other techniques which aim to either control the Lipschitz constant of the model or ensure its monotonicity. We show how the algorithm was used to train a powerful, robust, and interpretable discriminator for heavy-flavor decays in the LHCb realtime data-processing system.

Quantum reservoir computing using arrays of Rydberg atoms
Rodrigo Araiza Bravo, Khadijeh Najafi, Xun Gao, Susanne F. Yelin
[ arXiv:2111.10956 ]
Theoretical Physics

Abstract Quantum computing promises to provide machine learning with computational advantages. However, noisy intermediate-scale quantum (NISQ) devices pose engineering challenges to realizing quantum machine learning (QML) advantages. Recently, a series of QML computational models inspired by the noise-tolerant dynamics on the brain have emerged as a means to circumvent the hardware limitations of NISQ devices. In this article, we introduce a quantum version of a recurrent neural network (RNN), a well-known model for neural circuits in the brain. Our quantum RNN (qRNN) makes use of the natural Hamiltonian dynamics of an ensemble of interacting spin-1/2 particles as a means for computation. In the limit where the Hamiltonian is diagonal, the qRNN recovers the dynamics of the classical version. Beyond this limit, we observe that the quantum dynamics of the qRNN provide it quantum computational features that can aid it in computation. To this end, we study a qRNN based on arrays of Rydberg atoms, and show that the qRNN is indeed capable of replicating the learning of several cognitive tasks such as multitasking, decision making, and long-term memory by taking advantage of several key features of this platform such as interatomic species interactions, and quantum many-body scars.

New limits on the light dark matter: proton cross section from the cosmic large-scale structure
Keir K. Rogers, Cora Dvorkin, Hiranya V. Peiris
Physical Review Letters, 2022, Volume 128, Article 171301 [ arXiv:2111.10386 ] published
Astrophysics

Abstract We set the strongest limits to-date on the velocity-independent dark matter (DM) - proton cross section σ for DM masses m=10keV to 100GeV, using large-scale structure traced by the Lyman-alpha forest: e.g., a 95% lower limit σ<6×10−30cm2, for m=100keV. Our results complement direct detection, which has limited sensitivity to sub-GeV DM. We use an emulator of cosmological simulations, combined with data from the smallest cosmological scales used to-date, to model and search for the imprint of primordial DM-proton collisions. Cosmological bounds are improved by up to a factor of 25.

Equivariant Contrastive Learning
Rumen Dangovski, Li Jing, Charlotte Loh, Seungwook Han, Akash Srivastava, Brian Cheung, Pulkit Agrawal, Marin Soljačić
International Conference on Learned Representations, 2022 [ arXiv:2111.00899 | code ] published
Foundational AI

Abstract In state-of-the-art self-supervised learning (SSL) pre-training produces semantically good representations by encouraging them to be invariant under meaningful transformations prescribed from human knowledge. In fact, the property of invariance is a trivial instance of a broader class called equivariance, which can be intuitively understood as the property that representations transform according to the way the inputs transform. Here, we show that rather than using only invariance, pre-training that encourages non-trivial equivariance to some transformations, while maintaining invariance to other transformations, can be used to improve the semantic quality of representations. Specifically, we extend popular SSL methods to a more general framework which we name Equivariant Self-Supervised Learning (E-SSL). In E-SSL, a simple additional pre-training objective encourages equivariance by predicting the transformations applied to the input. We demonstrate E-SSL's effectiveness empirically on several popular computer vision benchmarks. Furthermore, we demonstrate usefulness of E-SSL for applications beyond computer vision; in particular, we show its utility on regression problems in photonics science. We will release our code.

Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
Charlotte Loh, Thomas Christensen, Rumen Dangovski, Samuel Kim, Marin Soljačić
Nature Communications, 2022, Volume 13, Article 4223 [ arXiv:2110.08406 | code ] published
Theoretical Physics

Abstract Deep learning techniques have been increasingly applied to the natural sciences, e.g., for property prediction and optimization or material discovery. A fundamental ingredient of such approaches is the vast quantity of labelled data needed to train the model; this poses severe challenges in data-scarce settings where obtaining labels requires substantial computational or labor resources. Here, we introduce surrogate- and invariance-boosted contrastive learning (SIB-CL), a deep learning framework which incorporates three ``inexpensive'' and easily obtainable auxiliary information sources to overcome data scarcity. Specifically, these are: 1)~abundant unlabeled data, 2)~prior knowledge of symmetries or invariances and 3)~surrogate data obtained at near-zero cost. We demonstrate SIB-CL's effectiveness and generality on various scientific problems, e.g., predicting the density-of-states of 2D photonic crystals and solving the 3D time-independent Schrodinger equation. SIB-CL consistently results in orders of magnitude reduction in the number of labels needed to achieve the same network accuracies.

Challenges for Unsupervised Anomaly Detection in Particle Physics
Katherine Fraser, Samuel Homiller, Rashmish K. Mishra, Bryan Ostdiek, Matthew D. Schwartz
Journal of High Energy Physics, 2022, Volume 2022, Article Number 66 [ arXiv:2110.06948 ] published
Experimental Physics

Abstract Anomaly detection relies on designing a score to determine whether a particular event is uncharacteristic of a given background distribution. One way to define a score is to use autoencoders, which rely on the ability to reconstruct certain types of data (background) but not others (signals). In this paper, we study some challenges associated with variational autoencoders, such as the dependence on hyperparameters and the metric used, in the context of anomalous signal (top and W) jets in a QCD background. We find that the hyperparameter choices strongly affect the network performance and that the optimal parameters for one signal are non-optimal for another. In exploring the networks, we uncover a connection between the latent space of a variational autoencoder trained using mean-squared-error and the optimal transport distances within the dataset. We then show that optimal transport distances to representative events in the background dataset can be used directly for anomaly detection, with performance comparable to the autoencoders. Whether using autoencoders or optimal transport distances for anomaly detection, we find that the choices that best represent the background are not necessarily best for signal identification. These challenges with unsupervised anomaly detection bolster the case for additional exploration of semi-supervised or alternative approaches.

A neural simulation-based inference approach for characterizing the Galactic Center γ-ray excess
Siddharth Mishra-Sharma, Kyle Cranmer
Physical Review D, 2922, Volume 105, Article 063017 [ arXiv:2110.06931 | code ] published
Astrophysics

Abstract The nature of the Fermi gamma-ray Galactic Center Excess (GCE) has remained a persistent mystery for over a decade. Although the excess is broadly compatible with emission expected due to dark matter annihilation, an explanation in terms of a population of unresolved astrophysical point sources e.g., millisecond pulsars, remains viable. The effort to uncover the origin of the GCE is hampered in particular by an incomplete understanding of diffuse emission of Galactic origin. This can lead to spurious features that make it difficult to robustly differentiate smooth emission, as expected for a dark matter origin, from more "clumpy" emission expected for a population of relatively bright, unresolved point sources. We use recent advancements in the field of simulation-based inference, in particular density estimation techniques using normalizing flows, in order to characterize the contribution of modeled components, including unresolved point source populations, to the GCE. Compared to traditional techniques based on the statistical distribution of photon counts, our machine learning-based method is able to utilize more of the information contained in a given model of the Galactic Center emission, and in particular can perform posterior parameter estimation while accounting for pixel-to-pixel spatial correlations in the gamma-ray map. This makes the method demonstrably more resilient to certain forms of model misspecification. On application to Fermi data, the method generically attributes a smaller fraction of the GCE flux to unresolved point sources when compared to traditional approaches. We nevertheless infer such a contribution to make up a non-negligible fraction of the GCE across all analysis variations considered, with at least 38+9−19% of the excess attributed to unresolved points sources in our baseline analysis.

Mixture Model Auto-Encoders: Deep Clustering through Dictionary Learning
Alexander Lin, Andrew H. Song, Demba Ba
ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2022, pp. 3368-3372 [ arXiv:2110.04683 | code ] published
Foundational AI

Abstract State-of-the-art approaches for clustering high-dimensional data utilize deep auto-encoder architectures. Many of these networks require a large number of parameters and suffer from a lack of interpretability, due to the black-box nature of the auto-encoders. We introduce Mixture Model Auto-Encoders (MixMate), a novel architecture that clusters data by performing inference on a generative model. Derived from the perspective of sparse dictionary learning and mixture models, MixMate comprises several auto-encoders, each tasked with reconstructing data in a distinct cluster, while enforcing sparsity in the latent space. Through experiments on various image datasets, we show that MixMate achieves competitive performance compared to state-of-the-art deep clustering algorithms, while using orders of magnitude fewer parameters.

Observation of enhanced free-electron radiation from photonic flatband resonancesg
Yi Yang, Charles Roques-Carmes, Steven E. Kooi, Haoning Tang, Justin Beroz, Eric Mazur, Ido Kaminer, John D. Joannopoulos, Marin Soljačić
Nature, 2023 [ arXiv:2110.03550 ] published
Foundational AI

Abstract Flatbands emerge from a myriad of structures such as Landau levels, Lieb and Kagome lattices, linegraphs, and more recently moire superlattices. They enable unique properties including slow light in photonics, correlated phases in electronics, and supercollimation in both systems. Despite these intense parallel efforts, flatbands have never been shown to affect the core light-matter interaction between electrons and photons, which is limited by a dimensionality mismatch. Here, we reveal that a photonic flatband can overcome this mismatch between localized electrons and extended photons and thus remarkably boost their light-matter interaction. We design flatband resonances in a silicon-on-insulator photonic crystal slab to control and enhance the radiation emission from free electrons by tuning their trajectory and velocity. In particular, we record a 100-fold radiation enhancement from the conventional diffraction-enabled Smith-Purcell radiation, and show the potential of our approach to achieve 106-fold enhancements and beyond. The enhancement also enables us to perform polarization shaping of free electron radiation from multiple flatbands and demonstrate an approach to measure photonic bands via angle-resolved electron-beam measurements. Our results suggest flatbands as ideal test beds for strong light-electron interaction in various systems, with particular relevance for efficient and compact free-electron light sources and accelerators.

Pruning a restricted Boltzmann machine for quantum state reconstruction
Anna Golubeva, Roger G. Melko
Physical Review B, 2022, Volume 105, Article 125124 [ arXiv:2110.03676 ] published
Theoretical Physics

Abstract Restricted Boltzmann machines (RBMs) have proven to be a powerful tool for learning quantum wavefunction representations from qubit projective measurement data. Since the number of classical parameters needed to encode a quantum wavefunction scales rapidly with the number of qubits, the ability to learn efficient representations is of critical importance. In this paper we study magnitude-based pruning as a way to compress the wavefunction representation in an RBM, focusing on RBMs trained on data from the transverse-field Ising model in one dimension. We find that pruning can reduce the total number of RBM weights, but the threshold at which the reconstruction accuracy starts to degrade varies significantly depending on the phase of the model. In a gapped region of the phase diagram, the RBM admits pruning over half of the weights while still accurately reproducing relevant physical observables. At the quantum critical point however, even a small amount of pruning can lead to significant loss of accuracy in the physical properties of the reconstructed quantum state. Our results highlight the importance of tracking all relevant observables as their sensitivity varies strongly with pruning. Finally, we find that sparse RBMs are trainable and discuss how a successful sparsity pattern can be created without pruning.

Classical Shadows for Quantum Process Tomography on Near-term Quantum Computers
Ryan Levy, Di Luo, Bryan K. Clark
Physical Review Research, 2024, Volume 6, Issue 1 [ arXiv:2110.02965 ] published
Theoretical Physics

Abstract Quantum process tomography is a powerful tool for understanding quantum channels and characterizing properties of quantum devices. Inspired by recent advances using classical shadows in quantum state tomography [H.-Y. Huang, R. Kueng, and J. Preskill, Nat. Phys. 16, 1050 (2020).], we have developed ShadowQPT, a classical shadow method for quantum process tomography. We introduce two related formulations with and without ancilla qubits. ShadowQPT stochastically reconstructs the Choi matrix of the device allowing for an a-posteri classical evaluation of the device on arbitrary inputs with respect to arbitrary outputs. Using shadows we then show how to compute overlaps, generate all k-weight reduced processes, and perform reconstruction via Hamiltonian learning. These latter two tasks are efficient for large systems as the number of quantum measurements needed scales only logarithmically with the number of qubits. A number of additional approximations and improvements are developed including the use of a pair-factorized Clifford shadow and a series of post-processing techniques which significantly enhance the accuracy for recovering the quantum channel. We have implemented ShadowQPT using both Pauli and Clifford measurements on the IonQ trapped ion quantum computer for quantum processes up to n=4 qubits and achieved good performance.

Inferring dark matter substructure with astrometric lensing beyond the power spectrum
Siddharth Mishra-Sharma
[ arXiv:2110.01620 | code ]
Astrophysics

Abstract Astrometry -- the precise measurement of positions and motions of celestial objects -- has emerged as a promising avenue for characterizing the dark matter population in our Galaxy. By leveraging recent advances in simulation-based inference and neural network architectures, we introduce a novel method to search for global dark matter-induced gravitational lensing signatures in astrometric datasets. Our method based on neural likelihood-ratio estimation shows significantly enhanced sensitivity to a cold dark matter population and more favorable scaling with measurement noise compared to existing approaches based on two-point correlation statistics, establishing machine learning as a powerful tool for characterizing dark matter using astrometric data.

Physics-Augmented Learning: A New Paradigm Beyond Physics-Informed Learning
Ziming Liu, Yunyue Chen, Yuanqi Du, Max Tegmark
[ arXiv:2109.13901 ]
Foundational AI

Abstract Integrating physical inductive biases into machine learning can improve model generalizability. We generalize the successful paradigm of physics-informed learning (PIL) into a more general framework that also includes what we term physics-augmented learning (PAL). PIL and PAL complement each other by handling discriminative and generative properties, respectively. In numerical experiments, we show that PAL performs well on examples where PIL is inapplicable or inefficient.

Presenting Unbinned Differential Cross Section Results
Miguel Arratia, Anja Butter, Mario Campanelli, Vincent Croft, Aishik Ghosh, Dag Gillberg, Kristin Lohwasser, Bogdan Malaescu, Vinicius Mikuni, Benjamin Nachman, Juan Rojo, Jesse Thaler, Ramon Winterhalder
Journal of Instrumentation, 2022, Volume 17 [ arXiv:2109.13243 | code ] published
Experimental Physics

Abstract Machine learning tools have empowered a qualitatively new way to perform differential cross section measurements whereby the data are unbinned, possibly in many dimensions. Unbinned measurements can enable, improve, or at least simplify comparisons between experiments and with theoretical predictions. Furthermore, many-dimensional measurements can be used to define observables after the measurement instead of before. There is currently no community standard for publishing unbinned data. While there are also essentially no measurements of this type public, unbinned measurements are expected in the near future given recent methodological advances. The purpose of this paper is to propose a scheme for presenting and using unbinned results, which can hopefully form the basis for a community standard to allow for integration into analysis workflows. This is foreseen to be the start of an evolving community dialogue, in order to accommodate future developments in this field that is rapidly evolving.

Overcoming the Spectral Bias of Neural Value Approximation
Ge Yang, Anurag Ajay, Pulkit Agrawal
ICLR 2022 Conference Proceedings [ arXiv:2206.04672 | code ] published
Foundational AI

Abstract Value approximation using deep neural networks is at the heart of off-policy deep reinforcement learning, and is often the primary module that provides learning signals to the rest of the algorithm. While multi-layer perceptron networks are universal function approximators, recent works in neural kernel regression suggest the presence of a \textit{spectral bias}, where fitting high-frequency components of the value function requires exponentially more gradient update steps than the low-frequency ones. In this work, we re-examine off-policy reinforcement learning through the lens of kernel regression and propose to overcome such bias via a composite neural tangent kernel. With just a single line-change, our approach, the Fourier feature networks (FFN) produce state-of-the-art performance on challenging continuous control domains with only a fraction of the compute. Faster convergence and better off-policy stability also make it possible to remove the target network without suffering catastrophic divergences, which further reduces TD(0)'s estimation bias on a few tasks. Code and analysis available at https://geyang.github.io/ffn.

Machine-learning hidden symmetries
Ziming Liu, Max Tegmark
Physical Review Letters, 2022, 128, 180201 [ arXiv:2109.09721 ] published

Abstract We present an automated method for finding hidden symmetries, defined as symmetries that become manifest only in a new coordinate system that must be discovered. Its core idea is to quantify asymmetry as violation of certain partial differential equations, and to numerically minimize such violation over the space of all invertible transformations, parametrized as invertible neural networks. For example, our method rediscovers the famous Gullstrand-Painleve metric that manifests hidden translational symmetry in the Schwarzschild metric of non-rotating black holes, as well as Hamiltonicity, modularity and other simplifying traits not traditionally viewed as symmetries.

Deep Set Auto Encoders for Anomaly Detection in Particle Physics
Bryan Ostdiek
SciPost Physics, 2022, Vol. 12, Issue 1 [ arXiv:2109.01695 ] published
Theoretical Physics

Abstract There is an increased interest in model agnostic search strategies for physics beyond the standard model at the Large Hadron Collider. We introduce a Deep Set Variational Autoencoder and present results on the Dark Machines Anomaly Score Challenge. We find that the method attains the best anomaly detection ability when there is no decoding step for the network, and the anomaly score is based solely on the representation within the encoded latent space. This method was one of the top-performing models in the Dark Machines Challenge, both for the open data sets as well as the blinded data sets.

Machine-Learning media bias
Samantha D’Alonzo, Max Tegmark
PLOS ONE, 2022, Volume 17, Issue 8, Article e0271947 [ arXiv:2109.00024 ] published
Foundational AI

Abstract We present an automated method for measuring media bias. Inferring which newspaper published a given article, based only on the frequencies with which it uses different phrases, leads to a conditional probability distribution whose analysis lets us automatically map newspapers and phrases into a bias space. By analyzing roughly a million articles from roughly a hundred newspapers for bias in dozens of news topics, our method maps newspapers into a two-dimensional bias landscape that agrees well with previous bias classifications based on human judgement. One dimension can be interpreted as traditional left-right bias, the other as establishment bias. This means that although news bias is inherently political, its measurement need not be.

What You Can Learn by Staring at a Blank Wall
Prafull Sharma, Miika Aittala, Yoav Y. Schechner, Antonio Torralba, Gregory W. Wornell, William T. Freeman, Fredo Durand
[ arXiv:2108.13027 | code ]
Foundational AI

Abstract We present a passive non-line-of-sight method that infers the number of people or activity of a person from the observation of a blank wall in an unknown room. Our technique analyzes complex imperceptible changes in indirect illumination in a video of the wall to reveal a signal that is correlated with motion in the hidden part of a scene. We use this signal to classify between zero, one, or two moving people, or the activity of a person in the hidden scene. We train two convolutional neural networks using data collected from 20 different scenes, and achieve an accuracy of ≈94% for both tasks in unseen test environments and real-time online settings. Unlike other passive non-line-of-sight methods, the technique does not rely on known occluders or controllable light sources, and generalizes to unknown rooms with no re-calibration. We analyze the generalization and robustness of our method with both real and synthetic data, and study the effect of the scene parameters on the signal quality.

Hardware-accelerated Inference for Real-Time Gravitational-Wave Astronomy
Alec Gunny, Dylan Rankin, Jeffrey Krupa, Muhammed Saleem, Tri Nguyen, Michael Coughlin, Philip Harris, Erik Katsavounidis, Steven Timm, Burt Holzman
[ arXiv:2108.12430 | code ]
Astrophysics

Abstract The field of transient astronomy has seen a revolution with the first gravitational-wave detections and the arrival of multi-messenger observations they enabled. Transformed by the first detection of binary black hole and binary neutron star mergers, computational demands in gravitational-wave astronomy are expected to grow by at least a factor of two over the next five years as the global network of kilometer-scale interferometers are brought to design sensitivity. With the increase in detector sensitivity, real-time delivery of gravitational-wave alerts will become increasingly important as an enabler of multi-messenger followup. In this work, we report a novel implementation and deployment of deep learning inference for real-time gravitational-wave data denoising and astrophysical source identification. This is accomplished using a generic Inference-as-a-Service model that is capable of adapting to the future needs of gravitational-wave data analysis. Our implementation allows seamless incorporation of hardware accelerators and also enables the use of commercial or private (dedicated) as-a-service computing. Based on our results, we propose a paradigm shift in low-latency and offline computing in gravitational-wave astronomy. Such a shift can address key challenges in peak-usage, scalability and reliability, and provide a data analysis platform particularly optimized for deep learning applications. The achieved sub-millisecond scale latency will also be relevant for any machine learning-based real-time control systems that may be invoked in the operation of near-future and next generation ground-based laser interferometers, as well as the front-end collection, distribution and processing of data from such instruments.

Real-time lattice gauge theory actions: unitarity, convergence, and path integral contour deformations
Gurtej Kanwar, Michael L. Wagman
Physical Review D, Volume 104, Article 014513 [ arXiv:2103.02602 ] published
Theoretical Physics

Abstract The Wilson action for Euclidean lattice gauge theory defines a positive-definite transfer matrix that corresponds to a unitary lattice gauge theory time-evolution operator if analytically continued to real time. Hoshina, Fujii, and Kikukawa (HFK) recently pointed out that applying the Wilson action discretization to continuum real-time gauge theory does not lead to this, or any other, unitary theory and proposed an alternate real-time lattice gauge theory action that does result in a unitary real-time transfer matrix. The character expansion defining the HFK action is divergent, and in this work we apply a path integral contour deformation to obtain a convergent representation for U(1) HFK path integrals suitable for numerical Monte Carlo calculations. We also introduce a class of real-time lattice gauge theory actions based on analytic continuation of the Euclidean heat-kernel action. Similar divergent sums are involved in defining these actions, but for one action in this class this divergence takes a particularly simple form, allowing construction of a path integral contour deformation that provides absolutely convergent representations for U(1) and SU(N) real-time lattice gauge theory path integrals. We perform proof-of-principle Monte Carlo calculations of real-time U(1) and SU(3) lattice gauge theory and verify that exact results for unitary time evolution of static quark-antiquark pairs in (1 + 1)D are reproduced.

Towards an Optimal Estimation of Cosmological Parameters with the Wavelet Scattering Transform
Georgios Valogiannis, Cora Dvorkin
Physical Review D, 2022, 105, 103534 [ arXiv:2108.07821 ] published
Astrophysics

Abstract Optimal extraction of the non-Gaussian information encoded in the Large-Scale Structure (LSS) of the universe lies at the forefront of modern precision cosmology. We propose achieving this task through the use of the Wavelet Scattering Transform (WST), which subjects an input field to a layer of non-linear transformations that are sensitive to non-Gaussianity in spatial density distributions through a generated set of WST coefficients. In order to assess its applicability in the context of LSS surveys, we apply the WST on the 3D overdensity field obtained by the Quijote simulations, out of which we extract the Fisher information in 6 cosmological parameters. It is subsequently found to deliver a large improvement in the marginalized errors on all parameters, ranging between 1.2−4× tighter than the corresponding ones obtained from the regular 3D cold dark matter + baryon power spectrum, as well as a 50% improvement over the neutrino mass constraint given by the marked power spectrum. Through this first application on 3D cosmological fields, we demonstrate the great promise held by this novel statistic and set the stage for its future application to actual galaxy observations.

A Deep-learning Approach for Live Anomaly Detection of Extragalactic Transients
Ashley Villar, Miles Cranmer, Edo Berger, Gabriella Contardo, Shirley Ho, Griffin Hosseinzadeh, Joshua Yao-Yu Lin
The Astrophysical Journal Supplement Series, Volume 255 [ | code ] published
Astrophysics

Abstract The Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo Interferometer Collaborations have now detected all three classes of compact binary mergers: binary black hole (BBH), binary neutron star (BNS), and neutron star-black hole (NSBH). For coalescences involving neutron stars, the simultaneous observation of gravitational and electromagnetic radiation produced by an event, has broader potential to enhance our understanding of these events, and also to probe the equation of state (EOS) of dense matter. However, electromagnetic follow-up to gravitational wave (GW) events requires rapid real-time detection and classification of GW signals, and conventional detection approaches are computationally prohibitive for the anticipated rate of detection of next-generation GW detectors. In this work, we present the first deep learning based results of classification of GW signals from NSBH mergers in extit{real} LIGO data. We show for the first time that a deep neural network can successfully distinguish all three classes of compact binary mergers and separate them from detector noise. Specifically, we train a convolutional neural network (CNN) on ∼500,000 data samples of real LIGO noise with injected BBH, BNS, and NSBH GW signals, and we show that our network has high sensitivity and accuracy. Most importantly, we successfully recover the two confirmed NSBH events to-date (GW200105 and GW200115) and the two confirmed BNS mergers to-date (GW170817 and GW190425), together with ≈90% of all BBH candidate events from the third Gravitational Wave Transient Catalog, GWTC-3. These results are an important step towards low-latency real-time GW detection, enabling multi-messenger astronomy.

Deep multi-task mining Calabi-Yau four-folds
Harold Erbin, Riccardo Finotello, Robin Schneider, Mohamed Tamaazousti
Machine Learning: Science and Technology, 2021, Volume 3, Number 1 [ arXiv:2108.02221 | code ] published
Theoretical Physics

Abstract We continue earlier efforts in computing the dimensions of tangent space cohomologies of Calabi-Yau manifolds using deep learning. In this paper, we consider the dataset of all Calabi-Yau four-folds constructed as complete intersections in products of projective spaces. Employing neural networks inspired by state-of-the-art computer vision architectures, we improve earlier benchmarks and demonstrate that all four non-trivial Hodge numbers can be learned at the same time using a multi-task architecture. With 30% (80%) training ratio, we reach an accuracy of 100% for h(1,1) and 97% for h(2,1) (100% for both), 81% (96%) for h(3,1), and 49% (83%) for h(2,2). Assuming that the Euler number is known, as it is easy to compute, and taking into account the linear constraint arising from index computations, we get 100% total accuracy.

Nonperturbative renormalization for the neural network–QFT correspondence
Harold Erbin, Vincent Lahoche, Dine Ousmane Samary
Machine Learning Science and Technology, 2022, Volume 3, Number 1, Article 015027 [ arXiv:2108.01403 | code ] published
Theoretical Physics

Abstract In a recent work~[1], Halverson, Maiti and Stoner proposed a description of neural networks in terms of a Wilsonian effective field theory. The infinite-width limit is mapped to a free field theory, while finite N corrections are taken into account by interactions (non-Gaussian terms in the action). In this paper, we study two related aspects of this correspondence. First, we comment on the concepts of locality and power-counting in this context. Indeed, these usual space-time notions may not hold for neural networks (since inputs can be arbitrary), however, the renormalization group provides natural notions of locality and scaling. Moreover, we comment on several subtleties, for example, that data components may not have a permutation symmetry: in that case, we argue that random tensor field theories could provide a natural generalization. Second, we improve the perturbative Wilsonian renormalization from~[1] by providing an analysis in terms of the nonperturbative renormalization group using the Wetterich-Morris equation. An important difference with usual nonperturbative RG analysis is that only the effective (IR) 2-point function is known, which requires setting the problem with care. Our aim is to provide a useful formalism to investigate neural networks behavior beyond the large-width limit (i.e.~far from Gaussian limit) in a nonperturbative fashion. A major result of our analysis is that changing the standard deviation of the neural network weight distribution can be interpreted as a renormalization flow in the space of networks. We focus on translations invariant kernels and provide preliminary numerical results.

Discovering Sparse Interpretable Dynamics from Partial Observations
Peter Y. Lu, Joan Ariño, Marin Soljačić
Communications Physics, 2022, Vol 5, Article 206 [ arXiv:2107.10879 | code ] published
Foundational AI

Abstract Identifying the governing equations of a nonlinear dynamical system is key to both understanding the physical features of the system and constructing an accurate model of the dynamics that generalizes well beyond the available data. We propose a machine learning framework for discovering these governing equations using only partial observations, combining an encoder for state reconstruction with a sparse symbolic model. Our tests show that this method can successfully reconstruct the full system state and identify the underlying dynamics for a variety of ODE and PDE systems.

QuanTaichi: A Compiler for Quantized Simulations
Yuanming Hu, Jiafeng Liu, Xuanda Yang, Mingkuan Xu, Ye Kuang, Weiwei Xu, Qiang Dai, William Freeman, Fredo Durand
ACM Transactions on Graphics, Volume 4, Article 182 [ | code ] published
Foundational AI

Abstract High-resolution simulations can deliver great visual quality, but they are often limited by available memory, especially on GPUs. We present a compiler for physical simulation that can achieve both high performance and significantly reduced memory costs, by enabling flexible and aggressive quantization. Low-precision ("quantized") numerical data types are used and packed to represent simulation states, leading to reduced memory space and bandwidth consumption. Quantized simulation allows higher resolution simulation with less memory, which is especially attractive on GPUs. Implementing a quantized simulator that has high performance and packs the data tightly for aggressive storage reduction would be extremely labor-intensive and error-prone using a traditional programming language. To make the creation of quantized simulation practical, we have developed a new set of language abstractions and a compilation system. A suite of tailored domain-specific optimizations ensure quantized simulators often run as fast as the full-precision simulators, despite the overhead of encoding-decoding the packed quantized data types. Our programming language and compiler, based on Taichi, allow developers to effortlessly switch between different full-precision and quantized simulators, to explore the full design space of quantization schemes, and ultimately to achieve a good balance between space and precision. The creation of quantized simulation with our system has large benefits in terms of memory consumption and performance, on a variety of hardware, from mobile devices to workstations with high-end GPUs. We can simulate with levels of resolution that were previously only achievable on systems with much more memory, such as multiple GPUs. For example, on a single GPU, we can simulate a Game of Life with 20 billion cells (8x compression per pixel), an Eulerian fluid system with 421 million active voxels (1.6x compression per voxel), and a hybrid Eulerian-Lagrangian elastic object simulation with 235 million particles (1.7x compression per particle). At the same time, quantized simulations create physically plausible results. Our quantization techniques are complementary to existing acceleration approaches of physical simulation: they can be used in combination with these existing approaches, such as sparse data structures, for even higher scalability and performance.

Neural Conditional Reweighting
Benjamin Nachman, Jesse Thaler
Physical Review D, Volume 105, Article 076015 [ arXiv:2107.08979 ] published
Theoretical Physics

Abstract There is a growing use of neural network classifiers as unbinned, high-dimensional (and variable-dimensional) reweighting functions. To date, the focus has been on marginal reweighting, where a subset of features are used for reweighting while all other features are integrated over. There are some situations, though, where it is preferable to condition on auxiliary features instead of marginalizing over them. In this paper, we introduce neural conditional reweighting, which extends neural marginal reweighting to the conditional case. This approach is particularly relevant in high-energy physics experiments for reweighting detector effects conditioned on particle-level truth information. We leverage a custom loss function that not only allows us to achieve neural conditional reweighting through a single training procedure, but also yields sensible interpolation even in the presence of phase space holes. As a specific example, we apply neural conditional reweighting to the energy response of high-energy jets, which could be used to improve the modeling of physics objects in parametrized fast simulation packages.

Flow-based sampling for multimodal distributions in lattice field theory
Daniel C. Hackett, Chung-Chun Hsieh, Michael S. Albergo, Denis Boyda, Jiunn-Wei Chen, Kai-Feng Chen, Kyle Cranmer, Gurtej Kanwar, Phiala E. Shanahan
[ arXiv:2107.00734 ]
Theoretical Physics

Abstract Recent results have demonstrated that samplers constructed with flow-based generative models are a promising new approach for configuration generation in lattice field theory. In this paper, we present a set of methods to construct flow models for targets with multiple separated modes (i.e. theories with multiple vacua). We demonstrate the application of these methods to modeling two-dimensional real scalar field theory in its symmetry-broken phase. In this context we investigate the performance of different flow-based sampling algorithms, including a composite sampling algorithm where flow-based proposals are occasionally augmented by applying updates using traditional algorithms like HMC.

Learning Task Informed Abstractions
Xiang Fu, Ge Yang, Pulkit Agrawal, Tommi Jaakkola
Proceedings of the 38th International Conference on Machine Learning, 2021, PMLR 139 [ arXiv:2106.15612 | code ] published
Foundational AI

Abstract Current model-based reinforcement learning methods struggle when operating from complex visual scenes due to their inability to prioritize task-relevant features. To mitigate this problem, we propose learning Task Informed Abstractions (TIA) that explicitly separates reward-correlated visual features from distractors. For learning TIA, we introduce the formalism of Task Informed MDP (TiMDP) that is realized by training two models that learn visual features via cooperative reconstruction, but one model is adversarially dissociated from the reward signal. Empirical evaluation shows that TIA leads to significant performance gains over state-of-the-art methods on many visual control tasks where natural and unconstrained visual distractions pose a formidable challenge.

The Principles of Deep Learning Theory
Daniel A. Roberts, Sho Yaida, Boris Hanin
Cambridge University Press (Book), 2022 [ arXiv:2106.10165 ] published
Foundational AI

Abstract This book develops an effective theory approach to understanding deep neural networks of practical relevance. Beginning from a first-principles component-level picture of networks, we explain how to determine an accurate description of the output of trained networks by solving layer-to-layer iteration equations and nonlinear learning dynamics. A main result is that the predictions of networks are described by nearly-Gaussian distributions, with the depth-to-width aspect ratio of the network controlling the deviations from the infinite-width Gaussian description. We explain how these effectively-deep networks learn nontrivial representations from training and more broadly analyze the mechanism of representation learning for nonlinear models. From a nearly-kernel-methods perspective, we find that the dependence of such models' predictions on the underlying learning algorithm can be expressed in a simple and universal way. To obtain these results, we develop the notion of representation group flow (RG flow) to characterize the propagation of signals through the network. By tuning networks to criticality, we give a practical solution to the exploding and vanishing gradient problem. We further explain how RG flow leads to near-universal behavior and lets us categorize networks built from different activation functions into universality classes. Altogether, we show that the depth-to-width ratio governs the effective model complexity of the ensemble of trained networks. By using information-theoretic techniques, we estimate the optimal aspect ratio at which we expect the network to be practically most useful and show how residual connections can be used to push this scale to arbitrary depths. With these tools, we can learn in detail about the inductive bias of architectures, hyperparameters, and optimizers.

Single electrons on solid neon as a solid-state quit platform
Xianjing Zhou, Gerwin Koolstra, Xufeng Zhang, Ge Yang, Xu Han, Brennan Dizdar, Divan Ralu, Wei Guo, Kater W. Murch, David I. Shuster, Dafei Jin
Nature, 2022, 605, 46-50 [ arXiv:2106.10326 ] published
Theoretical Physics

Abstract Progress toward the realization of quantum computers requires persistent advances in their constituent building blocks - qubits. Novel qubit platforms that simultaneously embody long coherence, fast operation, and large scalability offer compelling advantages in the construction of quantum computers and many other quantum information systems. Electrons, ubiquitous elementary particles of nonzero charge, spin, and mass, have commonly been perceived as paradigmatic local quantum information carriers. Despite superior controllability and configurability, their practical performance as qubits via either motional or spin states depends critically on their material environment. Here we report our experimental realization of a new qubit platform based upon isolated single electrons trapped on an ultraclean solid neon surface in vacuum. By integrating an electron trap in a circuit quantum electrodynamics architecture, we achieve strong coupling between the motional states of a single electron and a single microwave photon in an on-chip superconducting resonator. Qubit gate operations and dispersive readout are implemented to measure the energy relaxation time T1 of 15 μs and phase coherence time T2 over 200 ns. These results indicate that the electron-on-solid-neon qubit already performs near the state of the art as a charge qubit.

Toward Automatic Interpretation of 3D Plots
Laura E. Brandt, William T. Freeman
[ arXiv:2106.07627 ]
Foundational AI

Abstract This paper explores the challenge of teaching a machine how to reverse-engineer the grid-marked surfaces used to represent data in 3D surface plots of two-variable functions. These are common in scientific and economic publications; and humans can often interpret them with ease, quickly gleaning general shape and curvature information from the simple collection of curves. While machines have no such visual intuition, they do have the potential to accurately extract the more detailed quantitative data that guided the surface's construction. We approach this problem by synthesizing a new dataset of 3D grid-marked surfaces (SurfaceGrid) and training a deep neural net to estimate their shape. Our algorithm successfully recovers shape information from synthetic 3D surface plots that have had axes and shading information removed, been rendered with a variety of grid types, and viewed from a range of viewpoints.

Flow-based sampling for fermionic lattice field theories
Michael S. Albergo, Gurtej Kanwar, Sébastien Racanière, Danilo J. Rezende, Julian M. Urban, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Phiala E. Shanahan
Physical Review D, 2021, Vol. 104, Iss. 11 – 1 [ arXiv:2106.05934 ] published
Theoretical Physics

Abstract Algorithms based on normalizing flows are emerging as promising machine learning approaches to sampling complicated probability distributions in a way that can be made asymptotically exact. In the context of lattice field theory, proof-of-principle studies have demonstrated the effectiveness of this approach for scalar theories, gauge theories, and statistical systems. This work develops approaches that enable flow-based sampling of theories with dynamical fermions, which is necessary for the technique to be applied to lattice field theory studies of the Standard Model of particle physics and many condensed matter systems. As a practical demonstration, these methods are applied to the sampling of field configurations for a two-dimensional theory of massless staggered fermions coupled to a scalar field via a Yukawa interaction.

Light Field Networks: Neural Scene Representations with Single-Evaluation Rendering
Vincent Sitzmann, Semon Rezchikov, William T. Freeman, Joshua B. Tenenbaum, Fredo Durand
[ arXiv:2106.02634 | code ]
Foundational AI

Abstract Inferring representations of 3D scenes from 2D observations is a fundamental problem of computer graphics, computer vision, and artificial intelligence. Emerging 3D-structured neural scene representations are a promising approach to 3D scene understanding. In this work, we propose a novel neural scene representation, Light Field Networks or LFNs, which represent both geometry and appearance of the underlying 3D scene in a 360-degree, four-dimensional light field parameterized via a neural implicit representation. Rendering a ray from an LFN requires only a *single* network evaluation, as opposed to hundreds of evaluations per ray for ray-marching or volumetric based renderers in 3D-structured neural scene representations. In the setting of simple scenes, we leverage meta-learning to learn a prior over LFNs that enables multi-view consistent light field reconstruction from as little as a single image observation. This results in dramatic reductions in time and memory complexity, and enables real-time rendering. The cost of storing a 360-degree light field via an LFN is two orders of magnitude lower than conventional methods such as the Lumigraph. Utilizing the analytical differentiability of neural implicit representations and a novel parameterization of light space, we further demonstrate the extraction of sparse depth maps from LFNs.

Symmetry-via-Duality: Invariant Neural Network Densities from Parameter-Space Correlators
Anindita Maiti, Keegan Stoner, James Halverson
[ arXiv:2106.00694 | code ]
Theoretical Physics

Abstract Parameter-space and function-space provide two different duality frames in which to study neural networks. We demonstrate that symmetries of network densities may be determined via dual computations of network correlation functions, even when the density is unknown and the network is not equivariant. Symmetry-via-duality relies on invariance properties of the correlation functions, which stem from the choice of network parameter distributions. Input and output symmetries of neural network densities are determined, which recover known Gaussian process results in the infinite width limit. The mechanism may also be utilized to determine symmetries during training, when parameters are correlated, as well as symmetries of the Neural Tangent Kernel. We demonstrate that the amount of symmetry in the initialization density affects the accuracy of networks trained on Fashion-MNIST, and that symmetry breaking helps only when it is in the direction of ground truth.

Machine-Learning Non-Conservative Dynamics for New-Physics Detection
Ziming Liu, Bohan Wang, Qi Meng, Wei Chen, Max Tegmark, Tie-Yan Liu
Physical Review E, 2021, Vol. 104, Article 055302 [ arXiv:2106.00026 ] published

Abstract Energy conservation is a basic physics principle, the breakdown of which often implies new physics. This paper presents a method for data-driven "new physics" discovery. Specifically, given a trajectory governed by unknown forces, our Neural New-Physics Detector (NNPhD) aims to detect new physics by decomposing the force field into conservative and non-conservative components, which are represented by a Lagrangian Neural Network (LNN) and a universal approximator network (UAN), respectively, trained to minimize the force recovery error plus a constant λ times the magnitude of the predicted non-conservative force. We show that a phase transition occurs at λ=1, universally for arbitrary forces. We demonstrate that NNPhD successfully discovers new physics in toy numerical experiments, rediscovering friction (1493) from a damped double pendulum, Neptune from Uranus' orbit (1846) and gravitational waves (2017) from an inspiraling orbit. We also show how NNPhD coupled with an integrator outperforms previous methods for predicting the future of a damped double pendulum.

The Dark Machines Anomaly Score Challenge: Benchmark Data and Model Independent Event Classification for the Large Hadron Collider
T. Aarrestad, M. Van Beekveld, M. Bona, A. Bovenin, S. Caron, J. Davies, A. De Simone, C. Doglioni, J.M. Duarte, A. Farbin, H. Gupta, L. Hendriks, L. Heinrich, J. Howarth, P. Jawahar, A. Jueid, J. Lastow, A. Leinweber, J. Mamuzic, E. Merényi, A. Morandini, P. Moskvitina, C. Nellist, J. Ngadiuba, B. Ostdiek, M. Pierini, B. Ravina, R. Ruiz de Austri, S. Sekmen, M. Touranakou, M. Vaškevičiūte, R. Vilalta, J.-R. Vlimant, R. Verheyen, M. White, E. Wulff, E. Wallin, K.A. Wozniak, Z. Zhang
SciPost Physics, 2022, Volume 12, Issue 1, Page 43 [ arXiv:2105.14027 | code ] published
Experimental Physics

Abstract We describe the outcome of a data challenge conducted as part of the Dark Machines initiative and the Les Houches 2019 workshop on Physics at TeV colliders. The challenged aims at detecting signals of new physics at the LHC using unsupervised machine learning algorithms. First, we propose how an anomaly score could be implemented to define model-independent signal regions in LHC searches. We define and describe a large benchmark dataset, consisting of > 1 Billion simulated LHC events corresponding to 10 fb−1 of proton-proton collisions at a center-of-mass energy of 13 TeV. We then review a wide range of anomaly detection and density estimation algorithms, developed in the context of the data challenge, and we measure their performance in a set of realistic analysis environments. We draw a number of useful conclusions that will aid the development of unsupervised new physics searches during the third run of the LHC, and provide our benchmark dataset for future studies at https://www.phenoMLdata.org. Code to reproduce the analysis is provided at https://github.com/bostdiek/DarkMachines-UnsupervisedChallenge.

Covariance-Free Sparse Bayesian Learning
Alexander Lin, Andrew H. Song, Berkin Bilgic, Demba Ba
IEEE Transactions on Signal Processing, volume 70 [ arXiv:2105.10439 ] published
Foundational AI

Abstract Sparse Bayesian learning (SBL) is a powerful framework for tackling the sparse coding problem while also providing uncertainty quantification. The most popular inference algorithms for SBL exhibit prohibitively large computational costs for high-dimensional problems due to the need to maintain a large covariance matrix. To resolve this issue, we introduce a new method for accelerating SBL inference -- named covariance-free expectation maximization (CoFEM) -- that avoids explicit computation of the covariance matrix. CoFEM solves multiple linear systems to obtain unbiased estimates of the posterior statistics needed by SBL. This is accomplished by exploiting innovations from numerical linear algebra such as preconditioned conjugate gradient and a little-known diagonal estimation rule. For a large class of compressed sensing matrices, we provide theoretical justifications for why our method scales well in high-dimensional settings. Through simulations, we show that CoFEM can be up to thousands of times faster than existing baselines without sacrificing coding accuracy. Through applications to calcium imaging deconvolution and multi-contrast MRI reconstruction, we show that CoFEM enables SBL to tractably tackle high-dimensional sparse coding problems of practical interest.

Preserving New Physics while Simultaneously Unfolding All Observables
Patrick Komiske, W. Patrick McCormack, Benjamin Nachman
Physical Review D, Volume 104, Article 076027 [ arXiv:2105.09923 | code ] published
Theoretical Physics

Abstract Direct searches for new particles at colliders have traditionally been factorized into model proposals by theorists and model testing by experimentalists. With the recent advent of machine learning methods that allow for the simultaneous unfolding of all observables in a given phase space region, there is a new opportunity to blur these traditional boundaries by performing searches on unfolded data. This could facilitate a research program where data are explored in their natural high dimensionality with as little model bias as possible. We study how the information about physics beyond the Standard Model is preserved by full phase space unfolding using an important physics target at the Large Hadron Collider (LHC): exotic Higgs boson decays involving hadronic final states. We find that if the signal cross section is high enough, information about the new physics is visible in the unfolded data. We will show that in some cases, quantifiably all of the information about the new physics is encoded in the unfolded data. Finally, we show that there are still many cases when the unfolding does not work fully or precisely, such as when the signal cross section is small. This study will serve as an important benchmark for enhancing unfolding methods for the LHC and beyond.

Scaffolding Simulations with Deep Learning for High-dimensional Deconvolution
Anders Andreassen, Patrick T. Komiske, Eric M. Metodiev, Benjamin Nachman, Adi Suresh, and Jesse Thaler
Workshop paper at ICLR 2021 SimDL Workshop [ arXiv:2105.04448 | code ]

Abstract A common setting for scientific inference is the ability to sample from a high-fidelity forward model (simulation) without having an explicit probability density of the data. We propose a simulation-based maximum likelihood deconvolution approach in this setting called OmniFold. Deep learning enables this approach to be naturally unbinned and (variable-, and) high-dimensional. In contrast to model parameter estimation, the goal of deconvolution is to remove detector distortions in order to enable a variety of down-stream inference tasks. Our approach is the deep learning generalization of the common Richardson-Lucy approach that is also called Iterative Bayesian Unfolding in particle physics. We show how OmniFold can not only remove detector distortions, but it can also account for noise processes and acceptance effects.

A reconfigurable neural network ASIC for detector front-end data compression at the HL-LHC
Giuseppe Di Guglielmo, Farah Fahim, Christian Herwig, Manuel Blanco Valentin, Javier Duarte, Cristian Gingu, Philip Harris, James Hirschauer, Martin Kwok, Vladimir Loncar, Yingyi Luo, Llovizna Miranda, Jennifer Ngadiuba, Daniel Noonan, Seda Ogrenci-Memik, Maurizio Pierini, Sioni Summers, Nhan Tran
IEEE Transactions on Nuclear Science, 2021, Vol. 68, Issue 8 [ arXiv:2105.01683 ] published
Experimental Physics

Abstract Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains in the amount of data to be transported from the detector to off-detector logic where trigger decisions are made. We demonstrate that a neural network autoencoder model can be implemented in a radiation tolerant ASIC to perform lossy data compression alleviating the data transmission problem while preserving critical information of the detector energy profile. For our application, we consider the high-granularity calorimeter from the CMS experiment at the CERN Large Hadron Collider. The advantage of the machine learning approach is in the flexibility and configurability of the algorithm. By changing the neural network weights, a unique data compression algorithm can be deployed for each sensor in different detector regions, and changing detector or collider conditions. To meet area, performance, and power constraints, we perform a quantization-aware training to create an optimized neural network hardware implementation. The design is achieved through the use of high-level synthesis tools and the hls4ml framework, and was processed through synthesis and physical layout flows based on a LP CMOS 65 nm technology node. The flow anticipates 200 Mrad of ionizing radiation to select gates, and reports a total area of 3.6 mm^2 and consumes 95 mW of power. The simulated energy consumption per inference is 2.4 nJ. This is the first radiation tolerant on-detector ASIC implementation of a neural network that has been designed for particle physics applications.

Towards Designing and Exploiting Generative Networks for Neutrino Physics Experiments using Liquid Argon Time Projection Chambers
Paul Lutkus, Taritree Wongjirad, Schuchin Aeron
Conference paper at ICLR 2021 [ arXiv:2204.02496 | code ]
Experimental Physics

Abstract In this paper, we show that a hybrid approach to generative modeling via combin- ing the decoder from an autoencoder together with an explicit generative model for the latent space is a promising method for producing images of particle tra- jectories in a liquid argon time projection chamber (LArTPC). LArTPCs are a type of particle physics detector used by several current and future experiments focused on studies of the neutrino. We implement a Vector-Quantized Variational Autoencoder (VQ-VAE) and PixelCNN which produces images with LArTPC- like features and introduce a method to evaluate the quality of the images using a semantic segmentation that identifies important physics-based features.

Scalable and Flexible Deep Bayesian Optimization with Auxiliary Information for Scientific Problems
Samuel Kim, Peter Y. Lu, Charlotte Loh, Jamie Smith, Jasper Snoek, Marin Soljačić
Transactions on Machine Learning Research, September 2022 [ arXiv:2104.11667 | code ] published
Foundational AI

Abstract Bayesian optimization (BO) is a popular paradigm for global optimization of expensive black-box functions, but there are many domains where the function is not completely black-box. The data may have some known structure, e.g. symmetries, and the data generation process can yield useful intermediate or auxiliary information in addition to the value of the optimization objective. However, surrogate models traditionally employed in BO, such as Gaussian Processes (GPs), scale poorly with dataset size and struggle to incorporate known structure or auxiliary information. Instead, we propose performing BO on complex, structured problems by using Bayesian Neural Networks (BNNs), a class of scalable surrogate models that have the representation power and flexibility to handle structured data and exploit auxiliary information. We demonstrate BO on a number of realistic problems in physics and chemistry, including topology optimization of photonic crystal materials using convolutional neural networks, and chemical property optimization of molecules using graph neural networks. On these complex tasks, we show that BNNs often outperform GPs as surrogate models for BO in terms of both sampling efficiency and computational cost.

A Compound Poisson Generator approach to Point-Source Inference in Astrophysics
Gabriel H. Collin, Nicholas L. Rodd, Tyler Erjavec, Kerstin Perez
The Astrophysical Journal, 2022, Volume 260, Number 2 [ arXiv:2104.04529 | code ] published
Astrophysics

Abstract The identification and description of point sources is one of the oldest problems in astronomy; yet, even today the correct statistical treatment for point sources remains as one of the field's hardest problems. For dim or crowded sources, likelihood based inference methods are required to estimate the uncertainty on the characteristics of the source population. In this work, a new parametric likelihood is constructed for this problem using Compound Poisson Generator (CPG) functionals which incorporate instrumental effects from first principles. We demonstrate that the CPG approach exhibits a number advantages over Non-Poissonian Template Fitting (NPTF) - an existing parametric likelihood method - in a series of test scenarios in the context of X-ray astronomy. These demonstrations show that the effect of the point-spread function, effective area, and choice of point-source spatial distribution cannot, in general, be factorised as they are in the NPTF construction, while the new CPG construction is validated in these scenarios. Separately, an examination of the diffuse-flux emission limit is used to show that most simple choices of priors on the standard parameterisation of the population model can result in unexpected biases: when a model comprising both a point-source population and diffuse component is applied to this limit, nearly all observed flux will be assigned to either the population or to the diffuse component. A new parametrisation is presented for these priors which is demonstrated to properly estimate the uncertainties in this limit. In this choice of priors, the CPG correctly identifies that the fraction of flux assigned to the population model cannot be constrained by the data.

Why is AI hard and Physics simple?
Daniel A. Roberts
[ arXiv:2104.00008 ]
Foundational AI

Abstract We discuss why AI is hard and why physics is simple. We discuss how physical intuition and the approach of theoretical physics can be brought to bear on the field of artificial intelligence and specifically machine learning. We suggest that the underlying project of machine learning and the underlying project of physics are strongly coupled through the principle of sparsity, and we call upon theoretical physicists to work on AI as physicists. As a first step in that direction, we discuss an upcoming book on the principles of deep learning theory that attempts to realize this approach.

Machine Learning the 6th Dimension: Stellar Radial Velocities from 5D Phase-Space Correlations
Adriana Dropulic, Bryan Ostdiek, Laura J. Chang, Hongwan Liu, Timothy Cohen, and Mariangela Lisanti
The Astrophysical Journal Letters, 2021, 915, L14 [ arXiv:2103.14039 | code ] published
Astrophysics

Abstract The Gaia satellite will observe the positions and velocities of over a billion Milky Way stars. In the early data releases, the majority of observed stars do not have complete 6D phase-space information. In this Letter, we demonstrate the ability to infer the missing line-of-sight velocities until more spectroscopic observations become available. We utilize a novel neural network architecture that, after being trained on a subset of data with complete phase-space information, takes in a star's 5D astrometry (angular coordinates, proper motions, and parallax) and outputs a predicted line-of-sight velocity with an associated uncertainty. Working with a mock Gaia catalog, we show that the network can successfully recover the distributions and correlations of each velocity component for stars that fall within ∼5 kpc of the Sun. We also demonstrate that the network can accurately reconstruct the velocity distribution of a kinematic substructure in the stellar halo that is spatially uniform, even when it comprises a small fraction of the total star count.

Modern Machine Learning and Particle Physics
Matthew D. Schwartz
Harvard Data Science Review, 2021, Issue 3.2, 13 May [ arXiv:2103.12226 ] published
Theoretical Physics

Abstract Over the past five years, modern machine learning has been quietly revolutionizing particle physics. Old methodology is being outdated and entirely new ways of thinking about data are becoming commonplace. This article will review some aspects of the natural synergy between modern machine learning and particle physics, focusing on applications at the Large Hadron Collider. A sampling of examples is given, from signal/background discrimination tasks using supervised learning to direct data-driven approaches. Some comments on persistent challenges and possible future directions for the field are included at the end.

Deep learning: a statistical viewpoint
Peter L. Bartlett, Andrea Montanari, and Alexander Rakhlin
[ arXiv:2103.09177 ]
Foundational AI

Abstract The remarkable practical success of deep learning has revealed some major surprises from a theoretical perspective. In particular, simple gradient methods easily find near-optimal solutions to non-convex optimization problems, and despite giving a near-perfect fit to training data without any explicit effort to control model complexity, these methods exhibit excellent predictive accuracy. We conjecture that specific principles underlie these phenomena: that overparametrization allows gradient methods to find interpolating solutions, that these methods implicitly impose regularization, and that overparametrization leads to benign overfitting. We survey recent theoretical progress that provides examples illustrating these principles in simpler settings. We first review classical uniform convergence results and why they fall short of explaining aspects of the behavior of deep learning methods. We give examples of implicit regularization in simple settings, where gradient methods lead to minimal norm functions that perfectly fit the training data. Then we review prediction methods that exhibit benign overfitting, focusing on regression problems with quadratic loss. For these methods, we can decompose the prediction rule into a simple component that is useful for prediction and a spiky component that is useful for overfitting but, in a favorable setting, does not harm prediction accuracy. We focus specifically on the linear regime for neural networks, where the network can be approximated by a linear model. In this regime, we demonstrate the success of gradient flow, and we consider benign overfitting with two-layer networks, giving an exact asymptotic analysis that precisely demonstrates the impact of overparametrization. We conclude by highlighting the key challenges that arise in extending these insights to realistic deep learning settings.

hls4ml: An Open-Source Codesign Workflow to Empower Scientific Low-Power Machine Learning Devices
Farah Fahim, Benjamin Hawks, Christian Herwig, James Hirschauer, Sergo Jindariani, Nhan Tran, Luca P. Carloni, Giuseppe Di Guglielmo, Philip Harris, Jeffrey Krupa, Dylan Rankin, Manuel Blanco Valentin, Josiah Hester, Yingyi Luo, John Mamish, Seda Orgrenci-Memik, Thea Aarrestad, Hamza Javed, Vladimir Loncar, Maurizio Pierini, Adrian Alan Pol, Sioni Summers, Javier Duarte, Scott Hauck, Shih-Chieh Hsu, Jennifer Ngadiuba, Mia Liu, Duc Hoang, Edward Kreinar, Zhenbin Wu
[ arXiv:2103.05579 ]
Experimental Physics

Abstract Accessible machine learning algorithms, software, and diagnostic tools for energy-efficient devices and systems are extremely valuable across a broad range of application domains. In scientific domains, real-time near-sensor processing can drastically improve experimental design and accelerate scientific discoveries. To support domain scientists, we have developed hls4ml, an open-source software-hardware codesign workflow to interpret and translate machine learning algorithms for implementation with both FPGA and ASIC technologies. We expand on previous hls4ml work by extending capabilities and techniques towards low-power implementations and increased usability: new Python APIs, quantization-aware pruning, end-to-end FPGA workflows, long pipeline kernels for low power, and new device backends include an ASIC workflow. Taken together, these and continued efforts in hls4ml will arm a new generation of domain scientists with accessible, efficient, and powerful tools for machine-learning-accelerated discovery.

The Luminous and Double-Peaked Type Ic Supernova 2019stc: Evidence for Multiple Energy Sources
Sebastian Gomez, Edo Berger, Griffin Hosseinzadeh, Peter K. Blanchard, Matt Nicholl, V. Ashley Villar
The Astrophysical Journal, 2021, Vol. 913, Article 143 [ arXiv:2103.02611 ] published
Astrophysics

Abstract We present optical photometry and spectroscopy of SN\,2019stc (=ZTF19acbonaa), an unusual Type Ic supernova (SN Ic) at a redshift of z=0.117. SN\,2019stc exhibits a broad double-peaked light curve, with the first peak having an absolute magnitude of Mr=−20.0 mag, and the second peak, about 80 rest-frame days later, Mr=−19.2 mag. The total radiated energy is large, Erad≈2.5×1050 erg. Despite its large luminosity, approaching those of Type I superluminous supernovae (SLSNe), SN\,2019stc exhibits a typical SN Ic spectrum, bridging the gap between SLSNe and SNe Ic. The spectra indicate the presence of Fe-peak elements, but modeling of the first light curve peak with radioactive heating alone leads to an unusually high nickel mass fraction of fNi≈31% (MNi≈3.2 M⊙). Instead, if we model the first peak with a combined magnetar spin-down and radioactive heating model we find a better match with Mej≈4 M⊙, a magnetar spin period of Pspin≈7.2 ms and magnetic field of B≈1014 G, and fNi≲0.2 (consistent with SNe Ic). The prominent second peak cannot be naturally accommodated with radioactive heating or magnetar spin-down, but instead can be explained as circumstellar interaction with ≈0.7 M⊙ of hydrogen-free material located ≈400 AU from the progenitor. Including the remnant mass leads to a CO core mass prior to explosion of ≈6.5 M⊙. The host galaxy has a metallicity of ≈0.26 Z⊙, low for SNe Ic but consistent with SLSNe. Overall, we find that SN\,2019stc is a transition object between normal SNe Ic and SLSNe.

On the Minimal Error of Empirical Risk Minimization
Gil Kur, Alexander Rakhlin
[ arXiv:2102.12066 ]
Foundational AI

Abstract RWe study the minimal error of the Empirical Risk Minimization (ERM) procedure in the task of regression, both in the random and the fixed design settings. Our sharp lower bounds shed light on the possibility (or impossibility) of adapting to simplicity of the model generating the data. In the fixed design setting, we show that the error is governed by the global complexity of the entire class. In contrast, in random design, ERM may only adapt to simpler models if the local neighborhoods around the regression function are nearly as complex as the class itself, a somewhat counter-intuitive conclusion. We provide sharp lower bounds for performance of ERM for both Donsker and non-Donsker classes. We also discuss our results through the lens of recent studies on interpolation in overparameterized models.

Topological obstructions to autoencoding
Joshua Batson, C. Grace Haaf, Yonatan Kahn, Daniel A. Roberts
Journal of High Energy Physics, 2021, Issue 4, Article 280 [ arXiv:2102.08380 ] published
Theoretical Physics

Abstract Autoencoders have been proposed as a powerful tool for model-independent anomaly detection in high-energy physics. The operating principle is that events which do not belong to the space of training data will be reconstructed poorly, thus flagging them as anomalies. We point out that in a variety of examples of interest, the connection between large reconstruction error and anomalies is not so clear. In particular, for data sets with nontrivial topology, there will always be points that erroneously seem anomalous due to global issues. Conversely, neural networks typically have an inductive bias or prior to locally interpolate such that undersampled or rare events may be reconstructed with small error, despite actually being the desired anomalies. Taken together, these facts are in tension with the simple picture of the autoencoder as an anomaly detector. Using a series of illustrative low-dimensional examples, we show explicitly how the intrinsic and extrinsic topology of the dataset affects the behavior of an autoencoder and how this topology is manifested in the latent space representation during training. We ground this analysis in the discussion of a mock "bump hunt" in which the autoencoder fails to identify an anomalous "signal" for reasons tied to the intrinsic topology of n-particle phase space.

On the convergence of group-sparse autoencoders
Emmanouil Theodosis, Bahareh Tolooshams, Pranay Tankala, Abiy Tasissa, Demba Ba
[ arXiv:2102.07003 ]
Astrophysics

Abstract Recent approaches in the theoretical analysis of model-based deep learning architectures have studied the convergence of gradient descent in shallow ReLU networks that arise from generative models whose hidden layers are sparse. Motivated by the success of architectures that impose structured forms of sparsity, we introduce and study a group-sparse autoencoder that accounts for a variety of generative models, and utilizes a group-sparse ReLU activation function to force the non-zero units at a given layer to occur in blocks. For clustering models, inputs that result in the same group of active units belong to the same cluster. We proceed to analyze the gradient dynamics of a shallow instance of the proposed autoencoder, trained with data adhering to a group-sparse generative model. In this setting, we theoretically prove the convergence of the network parameters to a neighborhood of the generating matrix. We validate our model through numerical analysis and highlight the superior performance of networks with a group-sparse ReLU compared to networks that utilize traditional ReLUs, both in sparse coding and in parameter recovery tasks. We also provide real data experiments to corroborate the simulated results, and emphasize the clustering capabilities of structured sparsity models.

Few-nucleon matrix elements in pionless effective field theory in a finite volume
W. Detmold and P. E. Shanahan
Physical Review D, Volume 103, Article 074503 [ arXiv:2102.04329 ] published
Theoretical Physics

Abstract Pionless effective field theory in a finite volume (FVEFTπ/) is investigated as a framework for the analysis of multi-nucleon spectra and matrix elements calculated in lattice QCD (LQCD). By combining FVEFTπ/ with the stochastic variational method, the spectra of nuclei with atomic number A∈{2,3} are matched to existing finite-volume LQCD calculations at heavier-than-physical quark masses corresponding to a pion mass mπ=806 MeV, thereby enabling infinite-volume binding energies to be determined using infinite-volume variational calculations. Based on the variational wavefunctions that are constructed in this approach, the finite-volume matrix elements of various local operators are computed in FVEFTπ/ and matched to LQCD calculations of the corresponding QCD operators in the same volume, thereby determining the relevant one and two-body EFT counterterms and enabling an extrapolation of the LQCD matrix elements to infinite volume. As examples, the scalar, tensor, and axial matrix elements are considered, as well as the magnetic moments and the isovector longitudinal momentum fraction.

Path integral contour deformations for observables in SU(N) gauge theory
William Detmold, Gurtej Kanwar, Henry Lamm, Michael L. Wagman, Neill C. Warrington
Physical Review D, 2021, Vol. 103, Issue 9, Article 094517 [ arXiv:2101.12668 ] published
Theoretical Physics

Abstract Path integral contour deformations have been shown to mitigate sign and signal-to-noise problems associated with phase fluctuations in lattice field theories. We define a family of contour deformations applicable to SU(N) lattice gauge theory that can reduce sign and signal-to-noise problems associated with complex actions and complex observables. For observables, these contours can be used to define deformed observables with identical expectation value but different variance. As a proof-of-principle, we apply machine learning techniques to optimize the deformed observables associated with Wilson loops in two dimensional SU(2) and SU(3) gauge theory. We study loops consisting of up to 64 plaquettes and achieve variance reduction of up to 4 orders of magnitude.

The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics
Gregor Kasieczka (ed), Benjamin Nachman (ed), David Shih (ed), Oz Amram, Anders Andreassen, Kees Benkendorfer, Blaz Bortolato, Gustaaf Brooijmans, Florencia Canelli, Jack H. Collins, Biwei Dai, Felipe F. De Freitas, Barry M. Dillon, Ioan-Mihail Dinu, Zhongtian Dong, Julien Donini, Javier Duarte, D. A. Faroughy, Julia Gonski, Philip Harris, Alan Kahn, Jernej F. Kamenik, Charanjit K. Khosa, Patrick Komiske, Luc Le Pottier, Pablo Martín-Ramiro, Andrej Matevc, Eric Metodiev, Vinicius Mikuni, Inês Ochoa, Sang Eon Park, Maurizio Pierini, Dylan Rankin, Veronica Sanz, Nilai Sarda, Urous Seljak, Aleks Smolkovic, George Stein, Cristina Mantilla Suarez, Manuel Szewc, Jesse Thaler, Steven Tsan, Silviu-Marian Udrescu, Louis Vaslin, Jean-Roch Vlimant, Daniel Williams, Mikaeel Yunus
Reports on Progress in Physics, 2021, Volume 84, Number 12 [ arXiv:2101.08320 ] published
Experimental Physics

Abstract A new paradigm for data-driven, model-agnostic new physics searches at colliders is emerging, and aims to leverage recent breakthroughs in anomaly detection and machine learning. In order to develop and benchmark new anomaly detection methods within this framework, it is essential to have standard datasets. To this end, we have created the LHC Olympics 2020, a community challenge accompanied by a set of simulated collider events. Participants in these Olympics have developed their methods using an R&D dataset and then tested them on black boxes: datasets with an unknown anomaly (or not). This paper will review the LHC Olympics 2020 challenge, including an overview of the competition, a description of methods deployed in the competition, lessons learned from the experience, and implications for data analyses with future datasets as well as future colliders.

Introduction to Normalizing Flows for Lattice Field Theory
Michael S. Albergo, Denis Boyda, Daniel C. Hackett, Gurtej Kanwar, Kyle Cranmer, Sébastien Racanière, Danilo Jimenez Rezende, and Phiala E. Shanahan
[ arXiv:2101.08176 ]
Theoretical Physics

Abstract This notebook tutorial demonstrates a method for sampling Boltzmann distributions of lattice field theories using a class of machine learning models known as normalizing flows. The ideas and approaches proposed in arXiv:1904.12072, arXiv:2002.02428, and arXiv:2003.06413 are reviewed and a concrete implementation of the framework is presented. We apply this framework to a lattice scalar field theory and to U(1) gauge theory, explicitly encoding gauge symmetries in the flow-based approach to the latter. This presentation is intended to be interactive and working with the attached Jupyter notebook is recommended.

E Pluribus Unum Ex Machina: Learning from Many Collider Events at Once
Benjamin Nachman and Jesse Thaler
Physical Review D, 2021, Vol. 103, Issue 11, Article 116013 [ arXiv:2101.07263 | code ] published
Experimental Physics

Abstract There have been a number of recent proposals to enhance the performance of machine learning strategies for collider physics by combining many distinct events into a single ensemble feature. To evaluate the efficacy of these proposals, we study the connection between single-event classifiers and multi-event classifiers under the assumption that collider events are independent and identically distributed (IID). We show how one can build optimal multi-event classifiers from single-event classifiers, and we also show how to construct multi-event classifiers such that they produce optimal single-event classifiers. This is illustrated for a Gaussian example as well as for classification tasks relevant for searches and measurements at the Large Hadron Collider. We extend our discussion to regression tasks by showing how they can be phrased in terms of parametrized classifiers. Empirically, we find that training a single-event (per-instance) classifier is more effective than training a multi-event (per-ensemble) classifier, as least for the cases we studied, and we relate this fact to properties of the loss function gradient in the two cases. While we did not identify a clear benefit from using multi-event classifiers in the collider context, we speculate on the potential value of these methods in cases involving only approximate independence, as relevant for jet substructure studies.

Fast convolutional neural networks on FPGAs with hls4ml
Thea Aarrestad, Vladimir Loncar, Nicolò Ghielmetti, Maurizio Pierini, Sioni Summers, Jennifer Ngadiuba, Christoffer Petersson, Hampus Linander, Yutaro Iiyama, Giuseppe Di Guglielmo, Javier Duarte, Philip Harris, Dylan Rankin, Sergo Jindariani, Kevin Pedro, Nhan Tran, Mia Liu, Edward Kreinar, Zhenbin Wu, Duc Hoang
Machine Learning Science and Technology, 2021, Volume 2, Issue 4, Article 045015 [ arXiv:2101.05108 | code ] published
Experimental Physics

Abstract We introduce an automated tool for deploying ultra low-latency, low-power deep neural networks with convolutional layers on FPGAs. By extending the hls4ml library, we demonstrate an inference latency of 5μs using convolutional architectures, targeting microsecond latency applications like those at the CERN Large Hadron Collider. Considering benchmark models trained on the Street View House Numbers Dataset, we demonstrate various methods for model compression in order to fit the computational constraints of a typical FPGA device used in trigger and data acquisition systems of particle detectors. In particular, we discuss pruning and quantization-aware training, and demonstrate how resource utilization can be significantly reduced with little to no loss in model accuracy. We show that the FPGA critical resource consumption can be reduced by 97% with zero loss in model accuracy, and by 99% when tolerating a 6% accuracy degradation.

Detection and Parameter Estimation of Gravitational Waves from Binary Neutron-Star Mergers in Real LIGO Data using Deep Learning
Plamen G. Krastev, Kiranjyot Gill, V. Ashley Villar, Edo Berger
Physics Letters B, 2021, Vol. 815, Article 136161 [ arXiv:2012.13101 ] published
Astrophysics

Abstract One of the key challenges of real-time detection and parameter estimation of gravitational waves from compact binary mergers is the computational cost of conventional matched-filtering and Bayesian inference approaches. In particular, the application of these methods to the full signal parameter space available to the gravitational-wave detectors, and/or real-time parameter estimation is computationally prohibitive. On the other hand, rapid detection and inference are critical for prompt follow-up of the electromagnetic and astro-particle counterparts accompanying important transients, such as binary neutron-star and black-hole neutron-star mergers. Training deep neural networks to identify specific signals and learn a computationally efficient representation of the mapping between gravitational-wave signals and their parameters allows both detection and inference to be done quickly and reliably, with high sensitivity and accuracy. In this work we apply a deep-learning approach to rapidly identify and characterize transient gravitational-wave signals from binary neutron-star mergers in real LIGO data. We show for the first time that artificial neural networks can promptly detect and characterize binary neutron star gravitational-wave signals in real LIGO data, and distinguish them from noise and signals from coalescing black-hole binaries. We illustrate this key result by demonstrating that our deep-learning framework classifies correctly all gravitational-wave events from the Gravitational-Wave Transient Catalog, GWTC-1 [Phys. Rev. X 9 (2019), 031040]. These results emphasize the importance of using realistic gravitational-wave detector data in machine learning approaches, and represent a step towards achieving real-time detection and inference of gravitational waves.

Field of Junctions: Extracting Boundary Structure at Low SNR
Dor Verbin, Todd Zickler
IEEE/CVF International Conference on Computer Vision, 2021 [ arXiv:2011.13866 | code ] published
Foundational AI

Abstract We introduce a bottom-up model for simultaneously finding many boundary elements in an image, including contours, corners and junctions. The model explains boundary shape in each small patch using a 'generalized M-junction' comprising M angles and a freely-moving vertex. Images are analyzed using non-convex optimization to cooperatively find M+2 junction values at every location, with spatial consistency being enforced by a novel regularizer that reduces curvature while preserving corners and junctions. The resulting 'field of junctions' is simultaneously a contour detector, corner/junction detector, and boundary-aware smoothing of regional appearance. Notably, its unified analysis of contours, corners, junctions and uniform regions allows it to succeed at high noise levels, where other methods for segmentation and boundary detection fail.

AI Poincaré: Machine Learning Conservation Laws from Trajectories
Ziming Liu and Max Tegmark
Physical Review Letters, 2021, Volume 126, Issue 18, Article 180604 [ arXiv:2011.04698 ] published

Abstract We present AI Poincaré, a machine learning algorithm for auto-discovering conserved quantities using trajectory data from unknown dynamical systems. We test it on five Hamiltonian systems, including the gravitational 3-body problem, and find that it discovers not only all exactly conserved quantities, but also periodic orbits, phase transitions and breakdown timescales for approximate conservation laws.

Parameter Inference from Event Ensembles and the Top-Quark Mass
Forrest Flesher, Katherine Fraser, Charles Hutchison, Bryan Ostdiek, Matthew D. Schwartz
Journal of High Energy Physics, 2021, Article 58 [ arXiv:2011.04666 ] published

Abstract One of the key tasks of any particle collider is measurement. In practice, this is often done by fitting data to a simulation, which depends on many parameters. Sometimes, when the effects of varying different parameters are highly correlated, a large ensemble of data may be needed to resolve parameter-space degeneracies. An important example is measuring the top-quark mass, where other physical and unphysical parameters in the simulation must be marginalized over when fitting the top-quark mass parameter. We compare three different methodologies for top-quark mass measurement: a classical histogram fitting procedure, similar to one commonly used in experiment optionally augmented with soft-drop jet grooming; a machine-learning method called DCTR; and a linear regression approach, either using a least-squares fit or with a dense linearly-activated neural network. Despite the fact that individual events are totally uncorrelated, we find that the linear regression methods work most effectively when we input an ensemble of events sorted by mass, rather than training them on individual events. Although all methods provide robust extraction of the top-quark mass parameter, the linear network does marginally best and is remarkably simple. For the top study, we conclude that the Monte-Carlo-based uncertainty on current extractions of the top-quark mass from LHC data can be reduced significantly (by perhaps a factor of 2) using networks trained on sorted event ensembles. More generally, machine learning from ensembles for parameter estimation has broad potential for collider physics measurements.

Quasi Anomalous Knowledge: Searching for new physics with embedded knowledge
Sang Eon Park, Dylan Rankin, Silviu-Marian Udrescu, Mikaeel Yunus, Philip Harris
Journal of High Energy Physics, 2021, Article 30 [ arXiv:2011.03550 | code ] published
Experimental Physics

Abstract Discoveries of new phenomena often involve a dedicated search for a hypothetical physics signature. Recently, novel deep learning techniques have emerged for anomaly detection in the absence of a signal prior. However, by ignoring signal priors, the sensitivity of these approaches is significantly reduced. We present a new strategy dubbed Quasi Anomalous Knowledge (QUAK), whereby we introduce alternative signal priors that capture some of the salient features of new physics signatures, allowing for the recovery of sensitivity even when the alternative signal is incorrect. This approach can be applied to a broad range of physics models and neural network architectures. In this paper, we apply QUAK to anomaly detection of new physics events at the CERN Large Hadron Collider utilizing variational autoencoders with normalizing flow.

Learning to Unknot
Sergei Gukov, James Halverson, Fabian Ruehle, and Piotr Sułkowski
Machine Learning - Science and Technology, 2021, Volume 2, Number 2, Article 025035 [ arXiv:2010.16263 ] published
Theoretical Physics

Abstract We introduce natural language processing into the study of knot theory, as made natural by the braid word representation of knots. We study the UNKNOT problem of determining whether or not a given knot is the unknot. After describing an algorithm to randomly generate $N$-crossing braids and their knot closures and discussing the induced prior on the distribution of knots, we apply binary classification to the UNKNOT decision problem. We find that the Reformer and shared-QK Transformer network architectures outperform fully-connected networks, though all perform well. Perhaps surprisingly, we find that accuracy increases with the length of the braid word, and that the networks learn a direct correlation between the confidence of their predictions and the degree of the Jones polynomial. Finally, we utilize reinforcement learning (RL) to find sequences of Markov moves and braid relations that simplify knots and can identify unknots by explicitly giving the sequence of unknotting actions. Trust region policy optimization (TRPO) performs consistently well for a wide range of crossing numbers and thoroughly outperformed other RL algorithms and random walkers. Studying these actions, we find that braid relations are more useful in simplifying to the unknot than one of the Markov moves.

Mapping Machine-Learned Physics into a Human-Readable Space
Taylor Faucett, Jesse Thaler, Daniel Whiteson
Physics Review D, 2021, Volume 103, Iss. 3 [ arXiv:2010.11998 ] published
Experimental Physics

Abstract We present a technique for translating a black-box machine-learned classifier operating on a high-dimensional input space into a small set of human-interpretable observables that can be combined to make the same classification decisions. We iteratively select these observables from a large space of high-level discriminants by finding those with the highest decision similarity relative to the black box, quantified via a metric we introduce that evaluates the relative ordering of pairs of inputs. Successive iterations focus only on the subset of input pairs that are misordered by the current set of observables. This method enables simplification of the machine-learning strategy, interpretation of the results in terms of well-understood physical concepts, validation of the physical model, and the potential for new insights into the nature of the problem itself. As a demonstration, we apply our approach to the benchmark task of jet classification in collider physics, where a convolutional neural network acting on calorimeter jet images outperforms a set of six well-known jet substructure observables. Our method maps the convolutional neural network into a set of observables called energy flow polynomials, and it closes the performance gap by identifying a class of observables with an interesting physical interpretation that has been previously overlooked in the jet substructure literature.

Enhancing searches for resonances with machine learning and moment decomposition
Ouail Kitouni, Benjamin Nachman, Constantin Weisser, and Mike Williams
Journal of High Energy Physics, 2021, Article 70 [ arXiv:2010.09745 | code ] published
Experimental Physics

Abstract A key challenge in searches for resonant new physics is that classifiers trained to enhance potential signals must not induce localized structures. Such structures could result in a false signal when the background is estimated from data using sideband methods. A variety of techniques have been developed to construct classifiers which are independent from the resonant feature (often a mass). Such strategies are sufficient to avoid localized structures, but are not necessary. We develop a new set of tools using a novel moment loss function (Moment Decomposition or MoDe) which relax the assumption of independence without creating structures in the background. By allowing classifiers to be more flexible, we enhance the sensitivity to new physics without compromising the fidelity of the background estimation.

Elliptic stable envelopes and hypertoric loop spaces
Michael McBreen, Artan Sheshmani, Shing-Tung Yau
Selecta Mathematica, 2023, Volume 29, Article number 73 [ arXiv:2010.0067 ] published
Theoretical Physics

Abstract This paper relates the elliptic stable envelopes of a hypertoric variety X with the K-theoretic stable envelopes of the loop hypertoric space, ℒ˜X. It thus points to a possible categorification of elliptic stable envelopes.

AI Feynman: a Physics-Inspired Method for Symbolic Regression
Silviu-Marian Udrescu, Max Tegmark
Sciences Advances, 2020 [ arXiv:1905.11481 | code ] published
Foundational AI

Abstract A core challenge for both physics and artificial intellicence (AI) is symbolic regression: finding a symbolic expression that matches data from an unknown function. Although this problem is likely to be NP-hard in principle, functions of practical interest often exhibit symmetries, separability, compositionality and other simplifying properties. In this spirit, we develop a recursive multidimensional symbolic regression algorithm that combines neural network fitting with a suite of physics-inspired techniques. We apply it to 100 equations from the Feynman Lectures on Physics, and it discovers all of them, while previous publicly available software cracks only 71; for a more difficult test set, we improve the state of the art success rate from 15% to 90%.

Search for low mass vector resonances decaying into quark-antiquark pairs in proton-proton collisions at √𝑠 =13  TeV
CMS Collaboration
Physical Review D, 2019, Volume 100, Issue 11 [ arXiv:1909.04114 ] published
Experimental Physics

Abstract A search for low mass narrow vector resonances decaying into quark-antiquark pairs is presented. The analysis is based on data collected in 2017 with the CMS detector at the LHC in proton-proton collisions at a center-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 41.1 fb−1. The results of this analysis are combined with those of an earlier analysis based on data collected at the same collision energy in 2016, corresponding to 35.9 fb−1. Signal candidates will be recoiling against initial state radiation and are identified as energetic, large-radius jets with two pronged substructure. The invariant jet mass spectrum is probed for a potential narrow peaking signal over a smoothly falling background. No evidence for such resonances is observed within the mass range of 50-450 GeV. Upper limits at the 95% confidence level are set on the coupling of narrow resonances to quarks, as a function of the resonance mass. For masses between 50 and 300 GeV these are the most sensitive limits to date. This analysis extends the earlier search to a mass range of 300-450 GeV, which is probed for the first time with jet substructure techniques.

This project is supported by National Science Foundation under Cooperative Agreement PHY-2019786. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.