Solving Minimax Problems with Bilinear Objectives with ADMM

TL;DR

This paper presents an exact ADMM-based method for solving bilinear minimax problems by reducing the proximal operator to a generalized projection, enabling efficient saddle-point optimization.

Contribution

It introduces a novel exact reduction of the proximal operator to a generalized projection for bilinear saddle-point problems, simplifying ADMM implementation.

Findings

Proximal operator reduces to a generalized projection onto S.

The ADMM algorithm alternates between generalized projection and Euclidean projection.

The method guarantees convergence for bilinear minimax problems.

Abstract

We consider minimax (saddle-point) problems of the form max_{c \in C} min_{\beta \in S} g(c; \beta), where C and S are compact convex sets, and g is concave-convex. Applying the Alternating Direction Method of Multipliers (ADMM) requires evaluating a proximal operator that is, in general, as hard as the original problem. We show that when the outcome function g is bilinear, i.e. g(c; \beta) = c^T A \beta, the proximal operator reduces to a generalized projection onto the confidence region S. This reduction is exact -- it involves no approximation or linearization. The resulting ADMM algorithm alternates between (i) a generalized projection onto S and (ii) a Euclidean projection onto C. We describe the derivation, state the algorithm, and discuss convergence.