English

PRECISION: Decentralized Constrained Min-Max Learning with Low Communication and Sample Complexities

Machine Learning 2023-03-07 v1 Distributed, Parallel, and Cluster Computing

Abstract

Recently, min-max optimization problems have received increasing attention due to their wide range of applications in machine learning (ML). However, most existing min-max solution techniques are either single-machine or distributed algorithms coordinated by a central server. In this paper, we focus on the decentralized min-max optimization for learning with domain constraints, where multiple agents collectively solve a nonconvex-strongly-concave min-max saddle point problem without coordination from any server. Decentralized min-max optimization problems with domain constraints underpins many important ML applications, including multi-agent ML fairness assurance, and policy evaluations in multi-agent reinforcement learning. We propose an algorithm called PRECISION (proximal gradient-tracking and stochastic recursive variance reduction) that enjoys a convergence rate of O(1/T)O(1/T), where TT is the maximum number of iterations. To further reduce sample complexity, we propose PRECISION+^+ with an adaptive batch size technique. We show that the fast O(1/T)O(1/T) convergence of PRECISION and PRECISION+^+ to an ϵ\epsilon-stationary point imply O(ϵ2)O(\epsilon^{-2}) communication complexity and O(mnϵ2)O(m\sqrt{n}\epsilon^{-2}) sample complexity, where mm is the number of agents and nn is the size of dataset at each agent. To our knowledge, this is the first work that achieves O(ϵ2)O(\epsilon^{-2}) in both sample and communication complexities in decentralized min-max learning with domain constraints. Our experiments also corroborate the theoretical results.

Keywords

Cite

@article{arxiv.2303.02532,
  title  = {PRECISION: Decentralized Constrained Min-Max Learning with Low Communication and Sample Complexities},
  author = {Zhuqing Liu and Xin Zhang and Songtao Lu and Jia Liu},
  journal= {arXiv preprint arXiv:2303.02532},
  year   = {2023}
}
R2 v1 2026-06-28T09:01:39.664Z