Stay tuned for updates about our 2025 IAIFI Summer School Join our mailing list to receive updates.
Summer School 2024
The mission of the IAIFI PhD Summer School is to leverage the expertise of IAIFI researchers, affiliates, and partners toward promoting education and workforce development.
- August 5–9, 2024
- MIT, Stata Center (32 Vassar Street, Cambridge, MA), Room 155
ApplyAgenda Lecturers Tutorial LeadsAccommodations Costs Sponsors FAQ Past Schools
About
The Institute for Artificial Intelligence and Fundamental Interactions (IAIFI) is enabling physics discoveries and advancing foundational AI through the development of novel AI approaches that incorporate first principles, best practices, and domain knowledge from fundamental physics. The Summer School will include lectures and events that AI + Physics, illustrate interdisciplinary research at the intersection AI and Physics, and encourage diverse global networking. Hands-on code-based tutorials that build on foundational lecture materials help students put theory into practice.
Apply
Applications are closed for the 2024 IAIFI Summer School.
Accommodations
Students for the Summer School have the option to reserve dorm rooms (at their own expense) at Boston University. Instructions for this will be provided to students upon acceptance.
Costs
- There is no registration fee for the Summer School. Students for the Summer School are expected to cover the cost of travel and boarding.
- Lunch each day, as well as coffee and snacks at breaks, will be provided during the Summer School, along with at least one dinner during the Summer School.
- Students who wish to stay for the IAIFI Summer Workshop will be able to book the same rooms through the weekend and the Workshop if they choose (at their own expense).
Lecturers
Topic: Representation/Manifold Learning
Lecturer: Melanie Weber, Assistant Professor of Applied Mathematics and of Computer Science, Harvard
Topic: Uncertainty Quantification/Simulation-Based Inference
Lecturer: Carol Cuesta-Lazaro, IAIFI Fellow
Topic: Physics-Motivated Optimization
Lecturer: Cengiz Pehlevan, Assistant Professor of Applied Mathematics & Kempner Institute Associate Faculty, Harvard
Topic: Generative Models
Lecturer: Gilles Louppe, Professor, University of Liège
Tutorial Leads
Topic: Representation/Manifold Learning
Tutorial Lead: Sokratis Trifinopoulos (with Thomas Harvey, Incoming IAIFI Fellow)
Topic: Uncertainty Quantification/Simulation-Based Inference
Tutorial Lead: Jessie Micallef, IAIFI Fellow
Topic: Physics-Motivated Optimization
Tutorial Lead: Alex Atanasov, PhD Student, Harvard
Topic: Generative Modeling
Tutorial Lead: Gaia Grosso, IAIFI Fellow
Agenda
This agenda is subject to change.
Monday, August 5, 2024
9:00–9:30 am ET
Welcome/Introduction
9:30 am–12:00 pm ET
Lecture 1: Deep generative models: A latent variable model perspective, Gilles Louppe
Abstract
Deep generative models are probabilistic models that can be used as simulators of the data. They are used to generate samples, perform inference, or encode complex priors. In this lecture, we will review the principles of deep generative models from the unified perspective of latent variable models, covering variational auto-encoders, diffusion models, latent diffusion models, and normalizing flows. We will discuss the principles of variational inference, the training of generative models, and the interpretation of the latent space. Selected applications from scientific domains will also be presented.12:00–1:00 pm ET
Lunch
1:00–3:30 pm ET
Tutorial 1: Deep generative models: A latent variable model perspective, Gaia Grosso
3:30–4:30 pm ET
Introduction to Quantum Reservoir Learning with QuEra, Pedro Lopes and Milan Kornjača
Abstract
Quantum machine learning has gained considerable attention as quantum technology advances, presenting a promising approach for efficiently learning complex data patterns. Despite this promise, most contemporary quantum methods require significant resources for variational parameter optimization and face issues with vanishing gradients, leading to experiments that are either limited in scale or lack potential for quantum advantage. To address this, we develop a general-purpose, gradient-free, and scalable quantum reservoir learning algorithm that harnesses the quantum dynamics of QuEra's Aquila to process data. Quantum reservoir learning on Aquila, achieves competitive performance across various categories of machine learning tasks, including binary and multi-class classification, as well as time series prediction. The QuEra team performed successful quantum machine leaning demonstration on up to 108 qubits, demonstrating the largest quantum machine learning experiment to date. We also observe comparative quantum kernel advantage in learning tasks by constructing synthetic datasets based on the geometric differences between generated quantum and classical data kernels. In this presentation we will cover the general methods utilized to run quantum reservoir computing in QuEra's neutral-atom analog hardware, providing an introduction for users to pursue new research directions.5:00–7:00 pm ET
Welcome Dinner
Tuesday, August 6, 2024
9:00–9:30 am ET
Lightning Talks
9:30 am–12:00 pm ET
Lecture 2: Geometric Machine Learning, Melanie Weber
Abstract
A recent surge of interest in exploiting geometric structure in data and models in machine learning has motivated the design of a range of geometric algorithms and architectures. This lecture will give an overview of this emerging research area and its mathematical foundation. We will cover topics at the intersection of Geometry and Machine Learning, including relevant tools from differential geometry and group theory, geometric representation learning, graph machine learning, and geometric deep learning.12:00–1:00 pm ET
Lunch
1:00–3:30 pm ET
Tutorial 2: Geometric Machine Learning, Sokratis Trifinopoulos
3:30–4:30 pm ET
Breakout Sessions with Days 1 and 2 Lecturers and Tutorial Leads
4:30–6:00 pm ET
Group work for hackathon
Wednesday, August 7, 2024
9:00–9:30 am ET
Lightning Talks
9:30 am–12:00 pm ET
Lecture 3: Scaling and renormalization in high-dimensional regression, Cengiz Pehlevan
Abstract
From benign overfitting in overparameterized models to rich power-law scalings in performance, simple ridge regression displays surprising behaviors sometimes thought to be limited to deep neural networks. This balance of phenomenological richness with analytical tractability makes ridge regression the model system of choice in high-dimensional machine learning. In this set of lectures, I will present a unifying perspective on recent results on ridge regression using the basic tools of random matrix theory and free probability, aimed at researchers with backgrounds in physics and deep learning. I will highlight the fact that statistical fluctuations in empirical covariance matrices can be absorbed into a renormalization of the ridge parameter. This “deterministic equivalence” allows us to obtain analytic formulas for the training and generalization errors in a few lines of algebra by leveraging the properties of the S-transform of free probability. From these precise asymptotics, we can easily identify sources of power-law scaling in model performance. In all models, the S-transform corresponds to the train-test generalization gap, and yields an analogue of the generalized-cross-validation estimator.12:00–1:00 pm ET
Lunch
1:00–3:30 pm ET
Tutorial 3: Physics-motivated optimization: Scaling and renormalization in high-dimensional regression, Alex Atanasov
3:30–4:30 pm ET
Career Panel
-
Moderator: Alex Gagliano, IAIFI Fellow
-
Carol Cuesta-Lazaro, IAIFI Fellow
-
Pedro Lopes, Quantum Advocate, QuEra Computing
-
Gilles Louppe, Professor, University of Liège
-
Anton Mazurenko, Researcher, PDT Partners
-
Cengiz Pehlevan, Assistant Professor of Applied Mathematics & Kempner Institute Associate Faculty, Harvard
-
Partha Saha, Distinguished Engineer, Data and AI Platform, Visa
-
Melanie Weber, Assistant Professor of Applied Mathematics and of Computer Science, Harvard
Thursday, August 8, 2024
4:30–5:30 pm ET
Networking Reception sponsored by PDT Partners
9:00–9:30 am ET
Lightning Talks
9:30 am–12:00 pm ET
Lecture 4: Simulation-Based Inference for Modern Scientific Discovery, Carol Cuesta-Lazaro
Abstract
In this lecture, we will explore the foundations and applications of simulation-based inference (SBI) methods, demonstrating their efficiency in solving inverse problems across scientific disciplines. We'll introduce key approaches to SBI, including Neural Likelihood Estimation (NLE), Neural Posterior Estimation (NPE), and Neural Ratio Estimation (NRE), while highlighting their connections to data compression, density estimation, and generative models. Throughout the lecture, we'll address practical considerations such as computational efficiency and scalability to high-dimensional problems. We'll also examine recent advancements in the field, such as sequential methods and differentiable calibration, empowering attendees with the knowledge to adapt and customize these algorithms to their specific research challenges and constraints.12:00–1:00 pm ET
Lunch
1:00–3:30 pm ET
Tutorial 4: Uncertainty Quantification, Jessie Micallef
3:30–4:30 pm ET
Breakout Sessions with Days 3, 4, and 5 Lecturers and Tutorial Leads (Optional)
4:30–5:30 pm ET
Group work for hackathon
5:30–10:00 pm ET
Social Event with IAIFI members
Details
5:30–7:00 pm ET: Picnic with IAIFI members7:00–10:00 pm ET: Movie Night with MIT OpenSpace, The Imitation Game
Friday, August 9, 2024
9:00–9:30 am ET
Lightning Talks
9:30 am–12:00 pm ET
Hackathon
Projects
Project details to come.12:00–1:00 pm ET
Lunch
1:00–2:30 pm ET
Hackathon
Projects
Project details to come.2:30–3:30 pm ET
Hackathon presentations
3:30–4:00 pm ET
Closing
Financial Supporters
The Summer School is funded primarily by support from the National Science Foundation under Cooperative Agreement PHY-2019786. Computing resources are provided by the NSF ACCESS program.
We extend a sincere thank you to the following financial supporters of this year’s IAIFI Summer School:
We are also grateful to Foundry for providing compute as a prize for the hackathon.
2024 Organizing Committee
- Fabian Ruehle, Chair (Northeastern University)
- Demba Ba (Harvard)
- Alex Gagliano (IAIFI Fellow)
- Di Luo (IAIFI Fellow)
- Polina Abratenko (Tufts)
- Owen Dugan (MIT)
- Sneh Pandya (Northeastern)
- Yidi Qi (Northeastern)
- Manos Theodosis (Harvard)
- Sokratis Trifinopoulos (MIT)
FAQ
- Who can apply to the Summer School? Any PhD students or early career researchers working at the intersection of physics and AI may apply to the summer school.
- What is the cost to attend the Summer School? There is no registration fee for the Summer School. Students for the Summer School are expected to cover the cost of travel and boarding.
- Is there funding available to support my attendance at the Summer School? IAIFI is covering the cost of the Summer School, including lunch each day. There is no support available for travel costs.
- If I come to the Summer School, can I also attend the Workshop? Yes! We encourage you to stay for the IAIFI Summer Workshop and you can stay in the dorms for both events if you choose (at your own expense). Information about the Summer Workshop will be provided in early 2024.
- Will the recordings of the lectures be available? We expect to share recordings of the lectures after the Summer School.
- Will there be an option for virtual attendance? Yes, there is an option for virtual attendance.
- How can I book a dorm for the IAIFI Summer SchooL? Please email mailto:iaifi-summer@mit.edu to get the link for booking dorms at Boston University.
Contact iaifi@mit.edu with questions.