Abstract:We propose a novel trust region method for solving a class of nonsmooth, nonconvex composite-type optimization problems. The approach embeds inexact semismooth Newton steps for finding zeros of a normal map-based stationarity measure for the problem in a trust region framework. Based on a new merit function and acceptance mechanism, global convergence and transition to fast local q-superlinear convergence are established under standard conditions. In addition, we verify that the proposed trust region globalization is compatible with the Kurdyka-Łojasiewicz inequality yielding finer convergence results. We further derive new normal map-based representations of the associated second-order optimality conditions that have direct connections to the local assumptions required for fast convergence. Finally, we study the behavior of our algorithm when the Hessian matrix of the smooth part of the objective function is approximated by BFGS updates. We successfully link the KL theory, properties of the BFGS approximations, and a Dennis-Mor{é}-type condition to show superlinear convergence of the quasi-Newton version of our method. Numerical experiments on sparse logistic regression, image compression, and a constrained log-determinant problem illustrate the efficiency of the proposed algorithm.

A Trust Region Algorithm with Adaptive Cubic Regularization Methods for Nonsmooth Convex Minimization

A new trust region algorithm for nonsmooth convex minimization

A TRUST-REGION METHOD FOR NONSMOOTH NONCONVEX OPTIMIZATION

A Trust Region Algorithm for Constrained Nonsmooth Optimization Problems

A Trust Region Algorithm Model with Radius Bounded below for Minimization of Locally Lipschitzian Functions

Effective nonmonotone trust region method based on a simple cubic model for unconstrained optimization problems

Gradient trust region algorithm with limited memory BFGS update for nonsmooth convex minimization

A Superlinearly Convergent ODE-type Trust Region Algorithm for Nonsmooth Nonlinear Equations

An Adaptive Nonmonotonic Trust Region Method with Curvilinear Searches

An improved nonmonotone adaptive trust region method

A New Trust Region Method With Adaptive Radius For Unconstrained Optimization

A trust region-type normal map-based semismooth Newton method for nonsmooth nonconvex composite optimization

A Regularizing Trust Region Algorithm for Nonlinear Ill-Posed Problems

A robust trust region method for general constrained optimization

An adaptive simple model trust region algorithm based on new weak secant equations

An Improved Non-Monotonic Adaptive Trust Region Algorithm for Unconstrained Optimization

A modified adaptive cubic regularization method for large-scale unconstrained optimization problem

A Trust Region Method with Regularized Barzilai-Borwein Step-Size for Large-Scale Unconstrained Optimization

A New Modified Nonmonotone Adaptive Trust Region Method for Unconstrained Optimization

Nonmonotone Adaptive Trust Region Method Based on Simple Conic Model for Unconstrained Optimization

A Nonmonotone Trust Region Algorithm For Unconstrained Nonsmooth Optimization