Riemannian Convex Potential Maps

                 Samuel Cohen * 1 Brandon Amos * 2 Yaron Lipman 2 3

            Abstract
   Modeling distributions on Riemannian manifolds
   is a crucial component in understanding non-
   Euclidean data that arises, e.g., in physics and
   geology. The budding approaches in this space
   are limited by representational and computational
   tradeoffs. We propose and study a class of flows
   that uses convex potentials from Riemannian opti-
   mal transport. T ...