Wasserstein proximal learning

7a507f364fce7547f94b9a5b4a072c87?s=47 Wuchen Li
October 21, 2019

Wasserstein proximal learning

In this talk, I will briefly review the calculus behind optimal transport and mean-field games. In particular, I present the Wasserstein proximal, a.k.a Jordan-Kindler-Otto scheme, with the Hopf-Lax formula in density space and Master equations (Big Mac) in mean-field games. We demonstrate the usefulness of Wasserstein proximal in learning tasks through both generalization and optimization.


Wuchen Li

October 21, 2019