Stochastic Resetting: A Non-Equilibrium Framework for Prediction, Inference and Design

Abstract

Stochastic resetting has evolved from a simple model of diffusive search acceleration into a general framework for predicting, inferring, and controlling stochastic dynamics far from equilibrium. Its defining features, i.e., the creation of non-equilibrium steady states and the acceleration of first-passage kinetics, are increasingly relevant across physical chemistry, from biological restart mechanisms to molecular simulations and colloidal experiments. We review the renewal theory underlying stochastic resetting and show how it enables prediction of reset dynamics from properties of the underlying process, while also allowing the latter to be inferred from the resetting-accelerated dynamics. We then discuss applications to state preparation, enhanced sampling, kinetic inference, and training and sampling of machine learning models. Finally, we review recent advances in adaptive resetting, environmental feedback, many-body dynamics, and thermodynamic costs of resetting. These developments establish new opportunities for controlling stochastic dynamics with resetting across theory, simulations, and experiments.