Bridging AI and Non-equilibrium Statistical Mechanics: From Synthesizing Images to Simulating Turbulence

Misha Chertkov, University of Arizona
MCS Seminar Graphic

Abstract: In today's discourse, Artificial Intelligence (AI) takes center stage, but interpretations of its power, origin, place and future vary widely. In this talk, I'll share my personal enthusiasm for AI and its profound connection to, as well as its ability to advance, Non-equilibrium Statistical Mechanics. We'll explore two key themes:

  1. "How Statistical Mechanics Empowers AI": We delve into score-based diffusion models, highlighting the need to efficiently deconstruct fast correlations during the reverse (denoising) phase. We introduce "U-Turn Diffusion," a novel technique merging forward, U-turn, and reverse processes for generating independent and identically distributed samples/images. (Based on


  2. "How Physics-Informed AI Transforms Statistical Hydrodynamics": In the realm of simulating Turbulence, we challenge the traditional Large Eddy Simulation (LES) approach by developing Lagrangian LES empowered by AI. Leveraging Machine Learning tools of AI we bridge the gap between Eulerian (fields) and Lagrangian (particles) perspectives on Turbulence. (Based on


Bio: Dr. Michael (Misha) Chertkov is a Professor of Mathematics and Chair of the Graduate Interdisciplinary Program in Applied Mathematics at the UArizona. His research spans mathematics, statistical mechanics and artificial intelligence, with a focus on applications in fluid mechanics, control of engineered systems, and bio-social systems. He earned his Ph.D. in physics from the Weizmann Institute of Science in 1996 and has authored approximately 300 research papers. Dr. Chertkov holds the title of Fellow in both the AAAS and the American Physical Society and is a Senior Member of IEEE.