An inner accelerated inexact proximal augmented Lagrangian method and its iteration-complexity
por Jefferson Divino Gonçalves de Melo (Universidade Federal de Goiás)
Abstract
In this talk, we discuss about an inexact proximal augmented Lagrangian IPAL method for solving nonconvex composite optimization problems with nonlinear K-convex constraints, i.e., the constraints are convex with respect to the partial order given by a closed convex cone K. Each iteration of this scheme consists of inexactly solving a proximal augmented Lagrangian subproblem by an accelerated composite gradient algorithm followed by a Lagrange multiplier update. Under some mild assumptions, it is shown that IPAL generates an approximate stationary solution of the constrained problem in O(1/eps^3) inner iterations, where eps>0 is a given tolerance. Some numerical experiments to illustrate the computational efficiency of the proposed method will be presented.
This is a joint work with Renato D.C. Monteiro and Weiwei Kong - Georgia Tech/USA