I will propose in this mini tutorial an overview of novel approaches to compute optimal transport between measures, using a blend of recent techniques. After starting with the basics, describing the Monge and Kantorovich problems, I will show how convex optimization (using for instance regularization), automatic differentiation and neural networks can all prove useful to approximate optimal transport at scale, for various cost structures, and in high dimensional regimes.
MT7: Computational Optimal Transport
Organizer: Marco Cuturi
Apple, Inc., U.S.
This talk was given at the 2022 SIAM Conference on Mathematics of Data Science in San Diego, California, U.S. Learn more about SIAM Conferences at https://www.siam.org/conferences/about-siam-conferences
