ABSTRACT
The 2024 Nobel Prize in Physics was awarded to John Hopfield and Geoffrey Hinton "for foundational discoveries and inventions that enable machine learning with artificial neural networks." This presentation explains why neural-network research can win such an award in physics. We will trace how the awardees transformed the Ising model, a cornerstone of statistical physics, into artificial neural networks via a series of modifications. Building on this link, Hinton devised the first practical method to train deep neural networks and demonstrated their exceptional capabilities. This breakthrough ignited a flourishing of deep learning research, leading to the development of even better training methods. As physics facilitated the advancement of deep learning, machine learning has, in turn, repaid the favor by solving numerous physics problems. Therefore, physics and machine learning are deeply interconnected. Through this lens, we argue that machine learning represents a natural extension of physics - a perspective consistent with Hopfield's definition of physics.
BIOGRAPHY
Prof. Ge Zhang is an Assistant Professor of the Department of Physics, City University of Hong Kong. He earned his Ph.D. from Princeton University in 2017 by studying computational statistical physics. His research now focuses on the intersection between statistical physics and machine learning. He has published 30 papers in physics journals, including Physical Review X and Proceedings of the National Academy of Sciences. He has also published a paper in the International Conference on Machine Learning.
Date & Time
Venue