Boulder School 2027: Machine Learning and Condensed Matter

June 28-July 30, 2027

Scientific Coordinators

Roger G. Melko (University of Waterloo & Perimeter Institute)
Ehsan Khatami (San Jose State University)
Juan Carrasquilla (ETH)

Director: Leo Radzihovsky (University of Colorado, Boulder)

Machine learning is rapidly reshaping condensed matter and quantum many-body physics, enabling new approaches to understanding strongly correlated electrons, atomic matter, quantum devices, and more. At the same time, the learning algorithms themselves present fascinating phenomena for physicists to study, from generalization, to emergence, to the striking scaling laws observed in modern large language models. This summer school will introduce graduate students to the growing exchange of ideas between machine learning and many-body physics. The program will provide foundations in modern machine learning theory and methods, and explore applications to superconductivity, magnetism, atomic systems, and quantum simulators, including both a data-driven as well as a variational simulation perspective. Lectures from physicists and machine learning researchers will highlight recent progress and the growing exchange of ideas between these fields. The goal is to equip students with practical tools for frontier problems in condensed matter physics, while also showing how ideas from statistical mechanics, quantum many-body physics, and field theory can help illuminate the learning problem itself.

Lecturers and topics

Introduction to machine learning theory & methods (July 5-9)

Roger Melko (Waterloo): The learning problem; data, machine and loss.
Yasaman Bahri (Google): Large-width neural networks, mean-field limits.
*Anna Dawid (Lieden): Foundations of ML for the many-body problem.

Data-driven learning in classical and quantum systems (July 12-16)

Yi-Zhuang You (UCSD):
Ehsan Khatami (San Jose): Learning and strongly correlated electrons.

Neural quantum states (July 19-23)

Christopher Roth (Flatiron/UBC): NQS for interaction fermions.
*Annabelle Bohrdt (Munich): Hubbard models and more.
*Liang Fu (MIT): Transformer wavefunctions for fermions
Juan Carrasquilla (Zurich): Hamiltonian-driven variational learining.

Frontiers in physics and artificial intelligence (July 26-30)

Agnes Valenti (Flatiron): Generalization, double descent.
Zohar Ringel (Hebrew/Perimeter): RG, neural scaling laws, complexity.
Miles Stoudenmire (Flatiron): Tensor networks & neural networks.
*Andrey Gromov (Meta): Grokking, emergence and scaling laws.

*To be confirmed