ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2104.08708
22
40

Complexity Lower Bounds for Nonconvex-Strongly-Concave Min-Max Optimization

18 April 2021
Haochuan Li
Yi Tian
Jingzhao Zhang
Ali Jadbabaie
ArXivPDFHTML
Abstract

We provide a first-order oracle complexity lower bound for finding stationary points of min-max optimization problems where the objective function is smooth, nonconvex in the minimization variable, and strongly concave in the maximization variable. We establish a lower bound of Ω(κϵ−2)\Omega\left(\sqrt{\kappa}\epsilon^{-2}\right)Ω(κ​ϵ−2) for deterministic oracles, where ϵ\epsilonϵ defines the level of approximate stationarity and κ\kappaκ is the condition number. Our analysis shows that the upper bound achieved in (Lin et al., 2020b) is optimal in the ϵ\epsilonϵ and κ\kappaκ dependence up to logarithmic factors. For stochastic oracles, we provide a lower bound of Ω(κϵ−2+κ1/3ϵ−4)\Omega\left(\sqrt{\kappa}\epsilon^{-2} + \kappa^{1/3}\epsilon^{-4}\right)Ω(κ​ϵ−2+κ1/3ϵ−4). It suggests that there is a significant gap between the upper bound O(κ3ϵ−4)\mathcal{O}(\kappa^3 \epsilon^{-4})O(κ3ϵ−4) in (Lin et al., 2020a) and our lower bound in the condition number dependence.

View on arXiv
Comments on this paper