Smoothing Path-Following Method for Nonsmooth,Nonconvex Constrained Minimization

摘要

A smoothing path-following method is presented for a class of nonsmooth, nonconvex optimization problems, where the objective function f is nonsmooth and nonconvex, and the feasible set X is convex. From any starting point, this method finds a Clarke stationary point of the optimization problem by following a continuous path that is a solution of a smoothing ordinary differential equation (ODE) defined by a smooth approximation of the optimization problem.