A Smoothed Augmented Lagrangian Method for Nonsmooth Deterministic Convex problems

Abstract

We consider resolving the nonsmooth convex optimization problem. We present a smoothed augmented Lagrangian framework for resolving such problems where the Lagrangian subproblems are $L_k$-smooth and $L_k$ is a known constant. By employing smoothed Augmented Lagrangian schemes for getting approximate solutions of such problems and a suitably defined dual update, rate statements are established for the primal sub-optimality and infeasibility.

Date
May 29, 2023 10:00 AM — Jun 3, 2030 11:00 AM
Location
Seattle, USA