The Back-And-Forth Method For Wasserstein Gradient Flows

Tuesday, September 13, 2022 - 1:25pm - 2:25pm
Walter Library 402

Wonjun Lee (University of Minnesota, Twin Cities)

We present a method to efficiently compute Wasserstein gradient flows. Our approach is based on a generalization of the back-and-forth method (BFM) introduced by Jacobs and Leger to solve optimal transport problems. We evolve the gradient flow by solving the dual problem to the JKO scheme. In general, the dual problem is much better behaved than the primal problem. This allows us to efficiently run large scale gradient flows simulations for a large class of internal energies including singular and non-convex energies.

Joint work with Matt Jacobs (Purdue University) and Flavien Leger (INRIA Paris)