2014-11-13
10:15 at HCI J 8

Stochastic origins of gradient flows: a general connection

Mark Peletier

Technische Universiteit Eindhoven, Netherlands

Although it has been known for some time that certain gradient-flow structures are related to the large deviations of a stochastic process, until recently we only understood this at the level of examples. In this lecture I will explain a general structure that gives rise to the following property: for every sequence of reversible stochastic processes with a large-deviations principle, the limiting equation is a generalized gradient flow that maps one-to-one to the large-deviations rate function. Therefore the large class of reversible stochastic processes generates a correspondingly large class of generalized gradient flows. This is joint work with Michiel Renger and Alexander Mielke (both WIAS).