UBC Mathematics Lecture Series: Marco Cuturi (Part II)

Optimal transport theory provides practitioners from statistics, imaging, graphics or machine learning with a very powerful toolbox to compare probability measures. These tools translate however in their original form into computational schemes that can become intractable or suffer from instability (such as non-differentiability or estimation bias). We will present in these two lectures how a few insights from optimization theory and in particular a careful regularization can result in tools that are considerably easier to implement, run faster because they can take advantage of parallel hardware and behave better from a statistical perspective. We will highlight applications from diverse areas, from graphics and brain imaging to text analysis and parametric estimation.

Biography:

Marco Cuturi is a leading researcher in optimal transport and its applications to machine learning and related areas. One of his recent results, so-called `entropy regularized optimal transport’ has opened a way to practically solve optimal transport in challenging settings such as when dealing with high dimensional data, enabling the application of OT theory to machine learning and other problems in data sciences. He is a professor of statistics at CREST/ENSAE, Université Paris Saclay. He received his Ph.D. in 11/2005 from the Ecole des Mines de Paris. He worked as a post-doctoral researcher at the Institute of Statistical Mathematics, Tokyo, between 2005 and 2007, in the financial industry until 2008, and in the ORFE department of Princeton University until 2010 as a lecturer. He was an associate professor at the Graduate School of Informatics of Kyoto University between 2010 and 2016.

Note to Participants: 

This is the second of a two part lecture. The first part will be given on Nov 30 2017, in ESB 2012 at the same time. Details for Part I are available here.