The Frank-Wolfe algorithm: a short introduction

Sebastian Pokutta
2023-11-29
Abstract:In this paper we provide an introduction to the Frank-Wolfe algorithm, a method for smooth convex optimization in the presence of (relatively) complicated constraints. We will present the algorithm, introduce key concepts, and establish important baseline results, such as e.g., primal and dual convergence. We will also discuss some of its properties, present a new adaptive step-size strategy as well as applications.
Optimization and Control
What problem does this paper attempt to address?