TALK: "Sinkhorn Divergences: Bridging the gap between Optimal Transport and MMD"

Speaker: Aude Genevay , Université Paris Dauphine

Date: Friday, May 03, 2019

Time: 11:00 AM to 12:00 PM

Public: Yes

Location: Room G370

Event Type: Seminar

Room Description: Building 32

Host: Justin Solomon

Contact: Mieke Moran, (617) 253-5817, mieke@csail.mit.edu

Relevant URL:

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: "Sinkhorn Divergences: Bridging the gap between Optimal Transport and MMD"

ABSTRACT: Sinkhorn Divergences, based on entropy-regularized OT, were first introduced by Cuturi in 2013 as a solution to the computational burden of OT. However, this family of losses actually interpolates between OT (no regularization) and MMD (infinite regularization). This interpolation property is also true in terms of sample complexity, and thus regularizing OT breaks its curse of dimensionality. We will illustrate these theoretical claims on a set of learning problems like learning a distribution from samples.
SHORT BIO: Aude received her PhD in Mathematics from Ecole Normale Supérieure and Université Paris Dauphine in March 2019. She worked under the supervision of Gabriel Peyré on the properties of the entropic regularization for optimal transport, and its applications to machine learning problems.

Research Areas:
Graphics & Vision

Impact Areas:

This event is not part of a series.

Created by Mieke Moran Email at Wednesday, May 01, 2019 at 8:18 PM.