Kavli Affiliate: Yi Zhou | First 5 Authors: Ziyi Chen, Qunwei Li, Yi Zhou, , | Summary: Standard gradient descent-ascent (GDA)-type algorithms can only find stationary points in nonconvex minimax optimization, which are far more sub-optimal than local minimax points. In this work, we develop GDA-type algorithms that globally converge to local minimax points in […]
Continue.. Finding Local Minimax Points via (Stochastic) Cubic-Regularized GDA: Global Convergence and Complexity