Faster Stochastic Algorithms for Minimax Optimization under Polyak-Łojasiewicz Conditions
Published in Advances in Neural Information Processing Systems 35 (NeurIPS 2022), 2022
This paper considers stochastic first-order algorithms for minimax optimization under Polyak-Łojasiewicz conditions. We propose SPIDER-GDA for solving the finite-sum problem and an accelerated algorithm for the ill-conditioned case, which achieves a better stochastic first-oreder oracle complexity