EURO 2025 Leeds
Abstract Submission

41. Barzilai-Borwein methods for nonlinear composite optimization

Invited abstract in session MB-35: Nonlinear Optimization Algorithms and Applications: 1 , stream Continuous and mixed-integer nonlinear programming: theory and algorithms.

Monday, 10:30-12:00
Room: Michael Sadler LG15

Authors (first author is the speaker)

1. Hongchao Zhang
Louisiana State University

Abstract

This talk introduces new gradient methods to solve nonlinear composite optimization using Barzilai-Borwein stepsizes as frequently as possible.
When the objective function is convex, the method proposes new step size rules without using line search. For nonconvex objective function, novel line search rules are designed based on natural generalizations of the step size rules derived from the convex case. Global convergence and convergence rate will be discussed for the proposed methods. Very promising preliminary numerical results will also be presented in the talk.

Keywords

Status: accepted


Back to the list of papers