Inspired by recent advances in machine learning, and in particular by the concept of learning an optimizer, we investigate a class of proximal primal-dual optimizers with a fixed amount of memory. We derive convergence criteria and find several sub-classes corresponding to classical optimization methods such as Chambolle-Pock and Douglas-Rachford methods. Finally, we discuss the choice of algorithm instances for concrete problems.

This presentation is part of Contributed Presentation “CP1 - Contributed session 1”

Authors:

Sebastian Banert (KTH - Royal Institute of Technology)

Jonas Adler (KTH Royal Institute of Technology)

Johan Karlsson (KTH - Royal Institute of Technology)

Ozan Öktem (KTH - Royal Institute of Technology)

Axel Ringh (KTH - Royal Institute of Technology)

Keywords:

convex optimization, machine learning, primal-dual algorithms, proximal algorithms