Mirror Descent-Type Algorithms for the Variational Inequality Problem with Functional Constraints

In three linesMirror descent-type algorithms for variational inequality problems with functional constraints. Proposed methods alternate between productive and non-productive steps based on constraint values, with optimal convergence rates for bounded monotone operators and Lipschitz convex constraints. Applicable to GANs, reinforcement learning, and adversarial training.

Summary generated by Claude — human-verified