A Newton-MR algorithm with complexity guarantees for nonconvex smooth unconstrained optimization

arXiv2208.07095MaRDI QIDQ6407871FDOQ6407871

Publication date: 15 August 2022

Abstract: In this paper, we consider variants of Newton-MR algorithm for solving unconstrained, smooth, but non-convex optimization problems. Unlike the overwhelming majority of Newton-type methods, which rely on conjugate gradient algorithm as the primary workhorse for their respective sub-problems, Newton-MR employs minimum residual (MINRES) method. Recently, it has been established that MINRES has inherent ability to detect non-positive curvature directions as soon as they arise and certain useful monotonicity properties will be satisfied before such detection. We leverage these recent results and show that our algorithms come with desirable properties including competitive first and second-order worst-case complexities. Numerical examples demonstrate the performance of our proposed algorithms.

This page was built for publication: A Newton-MR algorithm with complexity guarantees for nonconvex smooth unconstrained optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6407871)