Browse Teach Curate

Negative Momentum for Constrained Minimax Games

A momentum-buffer-update framework that carries the negative-momentum technique from unconstrained saddle-point optimization into constrained minimax games, providing a universal enhancement to classic game-solving algorithms with a proven convergence guarantee under an entropy regularizer. Students learn why reversing the sign of the momentum term can damp the rotational dynamics that slow convergence to equilibrium in min-max games, and how the method extends to both normal-form and extensive-form games.

Medibot Text

Rapid Learning in Constrained Minimax Games with Negative Momentum

In this paper, we delve into the utilization of the negative momentum technique in constrained minimax games. From an intuitive mechanical standpoint, we introduce a novel framework for momentum buff…