Title: Matrix concentration and sparsification
Speaker: Nick Harvey (University of British Columbia)
Abstract:
In the past 10 years, several research communities have studied problems of the following form:
Given a matrix A represented as a sum of certain positive semidefinite matrices, we would like to
approximately represent A as a weighted sum of fewer of those matrices. Spectral sparsification
of graphs is a special case of this problem. The problem is also closely related to some concentration
inequalities for randomly sampling matrices. This talk will survey various results on this topic.