There is a large theory for gradient flows and stochastic gradient flows with convex potentials – with many applications. On the other hand, there is no general theory for nonconvex potentials if the dimension of the state space becomes large (or infinite) and the behavior one can expect is very model dependent. I will present a perspective based on the continuous renormalization group flow that applies to models that arise in statistical mechanics and quantum field theory. This talk is based on joint works with Thierry Bodineau and Benoit Dagallier.

Speaker: Roland Bauerschmidt, Courant Intstitute