The Back-And-Forth Method For Wasserstein Gradient Flows
Data Science Seminar
Wonjun Lee (University of Minnesota, Twin Cities)
You may attend the talk either in person in Walter 402 or register via Zoom. Registration is required to access the Zoom webinar.
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)