Recent advances in probabilistic machine learning have brought renewed attention to fundamental concepts in non-equilibrium thermodynamics. This bridge extends both ways: accelerate sampling and estimation in non-equilibrium simulation, as well as improve controlling, regularizing and estimation in diffusion models. In this talk, I will illustrate how concepts from statistical mechanics/stochastic thermodynamics and statistical inference can be translated to each other. I will begin with a central problem in physical chemistry---the estimation of free energy---and demonstrate how recent computational advances enable efficient and scalable solutions. In the end, I will demonstrate these ideas naturally extend to improve density estimation, energy regularization, and inference-time control in diffusion models.
Back to Sampling, Inference, and Data-Driven Physical Modeling in Scientific Machine Learning
Copyright. All Rights Reserved.