Talk Title:

Theory and Algorithms for Non(-Clarke)-Regular Optimization Problems

Abstract:

Despite the growing interest in nonconvex and nonsmooth optimization problems among the continuous optimization community, much of the current research focuses on Clarke-regular objectives and constraints. This regularity leads to favorable computational properties, but limits the practical applicability of the algorithms developed. In this talk, we will examine the prevalence of non-regular optimization problems in modern data science caused by the complex composition of nonconvex and nonsmooth functions. Emphasis will be placed on the challenges posed by non-regularity in terms of variational analysis, algorithmic design, and statistical inference, as well as our recent efforts to overcome these challenges.

Location:

1174 Etcheverry Hall or Zoom