Printable PDF
Department of Mathematics,
University of California San Diego

****************************

Math 278C: Seminar in Optimization and Data Science

Prof. Ali Mohammad Nezhad

University of North Carolina

Real Algebraic Geometry and Higher-Order Interior-Point Methods

Abstract:

Semi-definite programming is a fundamental class of convex optimization problems with applications ranging from combinatorial optimization and control to polynomial optimization. Interior-point methods solve such problems by following a central path toward an optimal solution, and the geometry of this path plays an important role in understanding their convergence behavior.

In this talk, I will first give a brief introduction to semi-definite programming, some of its applications, and the central-path framework underlying interior-point methods. I will then discuss a semi-algebraic viewpoint on the central path and explain how analytic reparametrizations can be used to study its behavior near optimal solutions. This perspective leads naturally to higher-order interior-point methods and super-linear convergence results, including in settings where standard regularity assumptions such as strict complementarity may fail. Finally, I will discuss how these ideas extend beyond semi-definite programming to polynomial optimization and, more broadly, to optimization problems with o-minimal or tame structure.

Host: Jiawang Nie

November 25, 2026

3:00 PM

Zoom: https://ucsd.zoom.us/j/99545592483 (Meeting ID: 995 4559 2483 / Password: OPT278CFA2)

Research Areas

Optimization

****************************