Nonequilibrium Thermodynamics

Sampling is a fundamental part of scientific computation and modern AI applications. Modern advances in nonequilibrium thermodynamics, such as Jarzynski equality and Crooks fluctuation theorem, can be exploited to devise more efficient sampling algorithms. See for instance the work on stochastic normalizing flows and this review. In this project, the student will participate in devising novel sampling algorithms inspired by nonequilibrium processes, and applying them to physics applications including the simulation of lattice field theories. See this for a recent set of lecture notes.