Скачать или смотреть Three examples of easy non convex optimizations

Distinguished Lecture organized by IEEE Signal Processing Society Student Branch, IIT Kharagpur.

Speaker: Dr Ami Wiesel, Associate Professor, Rachel and Selim Benin School of Computer Science and Engineering, Hebrew University, Jerusalem, Israel

Date: 30-08-2021

Abstract: In this talk, we will present a few recent works on hard non-convex optimizations which can be globally solved under favourable conditions. First, we will consider structure robust covariance estimation. Asymptotically, we will show that any stationary point is globally optimal. Second, we will consider the fair principal component analysis and filter design. We will show that when the targets are orthogonal, any local minima is globally optimal. Third, we will consider regression problems with binary weights. We will show that their Lagrange relaxation has a unique local maximin point when the noise is small.