LeAP-SSN: A Semismooth Newton Method with Global Convergence Rates

TL;DR

LeAP-SSN is a globally convergent semismooth Newton method with adaptive regularization that guarantees convergence from arbitrary points and achieves fast local convergence under mild conditions, applicable to nonconvex and convex problems.

Contribution

It introduces a parameter-free, globally convergent semismooth Newton method with adaptive Levenberg--Marquardt regularization, bridging global guarantees and superlinear local convergence.

Findings

Achieves $ ext{O}(1/k)$ convergence for convex problems.

Attains $ ext{O}(1/ oot2 ext{k})$ convergence in nonconvex settings.

Demonstrates practical efficiency on imaging, contact mechanics, and machine learning problems.

Abstract

We propose LeAP-SSN (Levenberg--Marquardt Adaptive Proximal Semismooth Newton method), a semismooth Newton-type method with a simple, parameter-free globalisation strategy that guarantees convergence from arbitrary starting points in nonconvex settings to stationary points, and under a Polyak--Lojasiewicz condition, to a global minimum, in Hilbert spaces. The method employs an adaptive Levenberg--Marquardt regularisation for the Newton steps, combined with backtracking, and does not require knowledge of problem-specific constants. We establish global nonasymptotic rates: $\mathcal{O}(1/k)$ for convex problems in terms of objective values, $\mathcal{O}(1/\sqrt{k})$ under nonconvexity in terms of subgradients, and linear convergence under a Polyak--Lojasiewicz condition. The algorithm achieves superlinear convergence under mild semismoothness and Dennis--Mor\'e or partial smoothness conditions, even for non-isolated minimisers. By combining strong global guarantees with superlinear local rates in a fully parameter-agnostic framework, LeAP-SSN bridges the gap between globally convergent algorithms and the fast asymptotics of Newton's method. The practical efficiency of the method is illustrated on representative problems from imaging, contact mechanics, and machine learning.