Efficient random coordinate descent algorithms for large-scale structured nonconvex optimization

TL;DR

This paper introduces and analyzes new random coordinate descent algorithms tailored for large-scale nonconvex optimization problems with a specific structure, demonstrating their convergence properties and practical efficiency.

Contribution

It proposes novel algorithms for structured nonconvex problems, providing convergence analysis and showing improved performance over existing methods.

Findings

Algorithms converge to stationary points asymptotically.

Sublinear convergence rate in expectation for certain measures.

Local linear convergence under an error bound condition.

Abstract

In this paper we analyze several new methods for solving nonconvex optimization problems with the objective function formed as a sum of two terms: one is nonconvex and smooth, and another is convex but simple and its structure is known. Further, we consider both cases: unconstrained and linearly constrained nonconvex problems. For optimization problems of the above structure, we propose random coordinate descent algorithms and analyze their convergence properties. For the general case, when the objective function is nonconvex and composite we prove asymptotic convergence for the sequences generated by our algorithms to stationary points and sublinear rate of convergence in expectation for some optimality measure. Additionally, if the objective function satisfies an error bound condition we derive a local linear rate of convergence for the expected values of the objective function. We also present extensive numerical experiments for evaluating the performance of our algorithms in comparison with state-of-the-art methods.