EECS Seminar: Machine Learning Meets Statistical Physics: Breaking New Ground Together

Abstract: The broad adoption of data science and machine learning in domain sciences has led to exciting advancements in recent years. At the same time, the inherently statistical nature of machine learning algorithms has allowed them to leverage well-established methods from statistics and statistical physics. In this talk, I will explore the interplay between these two disciplines and how their synergy drives breakthroughs at the forefront of both fields. I will illustrate these with examples from existing literature and showcase two highlights from my recent research: (i) designing advanced statistical sampling methods to analyze neural network performance, and (ii) enabling materials inverse design.

Bio: Ying Wai Li is a staff scientist at Los Alamos National Laboratory, where she is team leader of the AI and Scientific Computing Team. She obtained her Ph.D. in physics from the University of Georgia in 2012. Subsequently, she was a postdoctoral researcher (2012-2014) and an R&D computational scientist (2014-2019) at the National Center for Computational Sciences of Oak Ridge National Laboratory. Ying Wai currently serves on the editorial board of Computer Physics Communications, and as adjunct associate professor at University of Georgia. Her research interests span computational statistical physics, condensed matter physics, algorithm design, high-performance computing, and machine learning for computer simulations and data analytics.