Wasserstein barycenters from a PDE perspective (Carlier, Université Paris Dauphine)

Titel: Wasserstein barycenters from a PDE perspective
Zeit: 09.06.2020, 11-12 Uhr
Ort: Online (contact marius.yamakou@fau.de to get the data for the VC) Lehrstuhl für Angewandte Analysis (Alexander von Humboldt-Professur)

Guillaume Carlier (Université Paris Dauphine) (Hompage)

 

The Wasserstein barycenter is a way to interpolate between several probability measures that has become quite popular in machine learning and statistics. Such objects are characterized by an obstacle problem for a system of Monge-Ampère equations. I will give some regularity results and describe a regularization from which one can obtain further results, including a central limit theorem.