banner

Overview

While Deep Learning’s potential in industry and the sciences is without question, a fundamental theoretical understanding of large artificial networks is thus far elusive.
There is growing evidence that tools from theoretical physics, in particular quantum field theory and statistical mechanics, will be crucial to bridge the gap between theory and practice.

As recognized by the 2024 Nobel Prize in Physics, the early stages of artificial intelligence development were heavily influenced by theoretical physics.
This mini-workshop aims to retrace the physical ideas behind early models and connect them with the statistical description of modern deep networks.
This event brings together researchers from the entire spectrum between machine learning practice and high-energy theory.
We hope to initiate fruitful collaboration between the different communities.

The talks up to tea time will be designed for a general physics audience.
During the subsequent discussion, we will deepen the discussed topics and find points of contact.

Information

Time: 10:00 - 17:00, Tuesday, March 11th, 2025
Location: Lecture Hall, Kavli IPMU
Registration: form

Program

10:00 - 10:05 Jun’ichi Yokoyama (director of IPMU): opening address
10:05 - 10:30 Jia Liu (director of CD3 - IPMU): opening address
10:30 - 11:00 Koji Hashimoto (Kyoto U): Introduction to MLPhys: Why AI x physics
11:00 - 11:45 Yoshiyuki Kabashima (U Tokyo): Hopfield model and beyond
11:45 - 12:30 Masatoshi Imada (U Tokyo): Boltzmann machine and its applications to materials science
12:30 - 14:00 Lunch
14:00 - 14:45 Simeon Hellerman, Leander Thiele, and Elisa Ferreira (Kavli IPMU)
15:00 - 15:30 Tea time
15:30 - 17:00 Discussion

Co-organized by MLPhys
Supported by MEXT -KAKENHI- Grant-in-Aid for Transformative Research Areas (A) “Foundation of Machine Learning Physics”