Bridging Non-equilibrium Simulation and Probabilistic Machine Learning

Explore the intersection of non-equilibrium thermodynamics and probabilistic machine learning in this 32-minute conference talk by Yuanqi Du from Cornell University, presented at IPAM's Sampling, Inference, and Data-Driven Physical Modeling in Scientific Machine Learning Workshop. Discover how concepts from statistical mechanics, stochastic thermodynamics, and statistical inference can be translated between disciplines to create mutual benefits. Learn about accelerating sampling and estimation in non-equilibrium simulation while simultaneously improving control, regularization, and estimation in diffusion models. Begin with the fundamental problem of free energy estimation in physical chemistry and see how recent computational advances enable efficient and scalable solutions. Understand how these theoretical foundations naturally extend to enhance density estimation, energy regularization, and inference-time control in diffusion models, demonstrating the powerful bidirectional relationship between physical modeling and machine learning methodologies.