A PVT-Type Algorithm for Minimizing a Nonsmooth Convex Function

Abstract

A general framework of the (parallel variable transformation) PVT-type algorithm, called the PVT-MYR algorithm, for minimizing a non-smooth convex function is proposed, via the Moreau-Yosida regularization. As a particular scheme of this framework an ε-scheme is also presented. The global convergence of this algorithm is given under the assumptions of strong convexity of the objective function and an ε-descent condition determined by an ε-forced function. An appendix stating the proximal point algorithm is recalled in the last section.