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.

Global Convergence of a Filter-Trust-region Algorithm for Solving Nonsmooth Equations.

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

A FILTER TRUST-REGION METHOD FOR UNCONSTRAINED OPTIMIZATION

Convergence Analysis of a Trust-Region Multidimensional Filter Method for Nonlinear Complementarity Problems

A Trust Region Algorithm for Constrained Nonsmooth Optimization Problems

A MODIFIED TRUST-REGION METHOD WITH GLOBAL CONVERGENCE FOR SYMMETRIC NONLINEAR EQUATIONS

A FILTER-TRUST-REGION METHOD FOR LC~1 UNCONSTRAINED OPTIMIZATION AND ITS GLOBAL CONVERGENCE

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

An Improved Trust Region Algorithm for Nonlinear Equations.

An Ode-Type Filter-Trust-Region Algorithm For Solving Nonlinear Programming

BFGS Trust-Region Method for Symmetric Nonlinear Equations

A New Trust Region Method for Nonlinear Equations

A Nonmonotone Trust Region Method for Solving Symmetric Nonlinear Equations

Trust Region Filter SQP Method for Nonlinear Programming: Local Convergence

A Filter Method for Nonlinear Semidefinite Programming with Global Convergence

A New Trust-Region Method with Line Search for Solving Symmetric Nonlinear Equations

A BFGS Trust-Region Method for Nonlinear Equations

On the Modified Trust Region Algorithm for Nonlinear Equations

A MODIFIED FILTER RETROSPECTIVE TRUST REGION METHOD FOR UNCONSTRAINED OPTIMIZATION

A Nonmonotone Adaptive Trust-Region Algorithm for Symmetric Nonlinear Equations

A Filter Trust Region Method for Solving Semi-Infinite Programming Problems