Thời gian: 15h00 - VN time, thứ năm, ngày 13/4/2023.

Online: (google meet) https://meet.google.com/yyb-zhod-hdy?authuser=3&hl=vi

Abstract: In the first part of the talk, we present a parametric family of polynomials for maximum likelihood estimation, with applications to supervised learning. Based on Weierstrass' theorem and Putinar's Positivstellensatz, we guarantee the convergence of our polynomial estimations for exact probability density functions under mild conditions. Moreover, we show that our black-box optimization problem is a convex program with semidefinite constraints. Next, we apply Boyd's primal-dual subgradient method to solve this program numerically. This is joint work with Jean-Bernard Lasserre, Victor Magron, and Srecko Durasinovic.