Date: Jul 03, 2014, time: 09:30
Abstract. Optimal transport theory is an efficient tool to construct change of variables between probability densities. However, when it comes to the regularity of these maps, one cannot hope to obtain regularity estimates that are uniform with respect to the dimension except in some very special cases (for instance, between uniformly log-concave densities).
In random matrix theory the densities involved (modeling the distribution of the eigenvalues) are pretty singular, so it seems hopeless to apply optimal transport theory in this context. However, ideas coming from optimal transport can still be used to construct approximate transport maps (i.e., maps which send a density onto another up to a small error) which enjoy regularity estimates that are uniform in the dimension. Such maps can then be used to show universality results for the distribution of eigenvalues in random matrices.
The aim of this talk is to give a self-contained presentation of these results.