A Unified Analysis of Extra-gradient and Optimistic Gradient Methods for Saddle Point Problems: Proximal Point Approach
This work provides a theoretical foundation for gradient-based optimization algorithms in machine learning, addressing saddle point problems with incremental improvements in analysis.
The paper tackles the analysis of Extra-gradient (EG) and Optimistic Gradient Descent Ascent (OGDA) methods for saddle point problems by unifying them as approximations of the proximal point method, enabling a new framework for analysis in bilinear and strongly convex-strongly concave settings and generalizing OGDA results for a wide parameter range.
In this paper we consider solving saddle point problems using two variants of Gradient Descent-Ascent algorithms, Extra-gradient (EG) and Optimistic Gradient Descent Ascent (OGDA) methods. We show that both of these algorithms admit a unified analysis as approximations of the classical proximal point method for solving saddle point problems. This viewpoint enables us to develop a new framework for analyzing EG and OGDA for bilinear and strongly convex-strongly concave settings. Moreover, we use the proximal point approximation interpretation to generalize the results for OGDA for a wide range of parameters.