Fast Decentralized Gradient Tracking for Federated Minimax Optimization with Local Updates

arXiv:2405.04566v1 Announce Type: new
Abstract: Federated learning (FL) for minimax optimization has emerged as a powerful paradigm for training models across distributed nodes/clients while preserving data privacy and model robustness on data heterogeneity. In this work, we delve into the decentralized implementation of federated minimax optimization by proposing \texttt{K-GT-Minimax}, a novel decentralized minimax optimization algorithm that combines local updates and gradient tracking techniques. Our analysis showcases the algorithm's communication efficiency and convergence rate for nonconvex-strongly-concave (NC-SC) minimax optimization, demonstrating a …

abstract arxiv cs.dc cs.lg data data privacy decentralized distributed federated learning gradient implementation minimax model robustness nodes optimization paradigm privacy robustness stat.ml tracking training training models type updates while work