Last-iterate convergence: zero-sum games and constrained min-max optimization

DOI10.4230/LIPICS.ITCS.2019.27MaRDI QIDQ5090401FDOQ5090401

Authors: Constantinos Daskalakis, Ioannis Panageas 

Publication date: 18 July 2022

Full work available at URL: https://arxiv.org/abs/1807.04252

Recommendations

• Convergence rate of \(\mathcal{O}(1/k)\) for optimistic gradient and extragradient methods in smooth convex-concave saddle point problems
• Last-iterate convergence of saddle-point optimizers via high-resolution differential equations
• Fast convergence of optimistic gradient ascent in network zero-sum extensive form games
• Alternating Proximal-Gradient Steps for (Stochastic) Nonconvex-Concave Minimax Problems

zbMATH Keywords

convergencedynamical systemszero-sum gamesKL divergenceno regret learning

Mathematics Subject Classification ID

Theory of computing (68Qxx)

Cites Work

• Prediction, Learning, and Games
• Smooth minimization of non-smooth functions
• Title not available (Why is that?)
• An analog of the minimax theorem for vector payoffs
• An iterative method of solving a game
• Discrete Dynamical Systems
• Title not available (Why is that?)
• The equivalence of linear programs and zero-sum games
• Mutation, Sexual Reproduction and Survival in Dynamic Environments

Cited In (4)

• Alleviating limit cycling in training GANs with an optimization technique
• Last-iterate convergence of saddle-point optimizers via high-resolution differential equations
• Efficient second-order optimization with predictions in differential games
• A unified stochastic approximation framework for learning in games

This page was built for publication: Last-iterate convergence: zero-sum games and constrained min-max optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5090401)