The landscape of the proximal point method for nonconvex–nonconcave minimax optimization

被引:0
|
作者
Benjamin Grimmer
Haihao Lu
Pratik Worah
Vahab Mirrokni
机构
[1] Johns Hopkins University,
[2] University of Chicago,undefined
[3] Google Research,undefined
来源
Mathematical Programming | 2023年 / 201卷
关键词
Minimax optimization; Proximal point method; Moreau envelope; Nonconvex; Nonconcave; Linear convergence; Cycling; 90C47; 65K15; 49K35;
D O I
暂无
中图分类号
学科分类号
摘要
Minimax optimization has become a central tool in machine learning with applications in robust optimization, reinforcement learning, GANs, etc. These applications are often nonconvex–nonconcave, but the existing theory is unable to identify and deal with the fundamental difficulties this poses. In this paper, we study the classic proximal point method (PPM) applied to nonconvex–nonconcave minimax problems. We find that a classic generalization of the Moreau envelope by Attouch and Wets provides key insights. Critically, we show this envelope not only smooths the objective but can convexify and concavify it based on the level of interaction present between the minimizing and maximizing variables. From this, we identify three distinct regions of nonconvex–nonconcave problems. When interaction is sufficiently strong, we derive global linear convergence guarantees. Conversely when the interaction is fairly weak, we derive local linear convergence guarantees with a proper initialization. Between these two settings, we show that PPM may diverge or converge to a limit cycle.
引用
收藏
页码:373 / 407
页数:34
相关论文
共 50 条
  • [1] The landscape of the proximal point method for nonconvex-nonconcave minimax optimization
    Grimmer, Benjamin
    Lu, Haihao
    Worah, Pratik
    Mirrokni, Vahab
    MATHEMATICAL PROGRAMMING, 2023, 201 (1-2) : 373 - 407
  • [2] PROXIMAL POINT ALGORITHMS FOR NONCONVEX-NONCONCAVE MINIMAX OPTIMIZATION PROBLEMS
    Li, Xiao-bing
    Jiang, Yuan-xin
    Yao, Bin
    JOURNAL OF NONLINEAR AND CONVEX ANALYSIS, 2024, 25 (08) : 2007 - 2021
  • [3] Universal Gradient Descent Ascent Method for Nonconvex-Nonconcave Minimax Optimization
    Zheng, Taoli
    Zhu, Linglingzhi
    So, Anthony Man-Cho
    Blanchet, José
    Li, Jiajin
    arXiv, 2022,
  • [4] Universal Gradient Descent Ascent Method for Nonconvex-Nonconcave Minimax Optimization
    Zheng, Taoli
    Zhu, Linglingzhi
    So, Anthony Man-Cho
    Blanchet, Jose
    Li, Jiajin
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,
  • [5] Universal Gradient Descent Ascent Method for Nonconvex-Nonconcave Minimax Optimization
    Zheng, Taoli
    Zhu, Linglingzhi
    So, Anthony Man-Cho
    Blanchet, José
    Li, Jiajin
    Advances in Neural Information Processing Systems, 2023, 36 : 54075 - 54110
  • [6] What is Local Optimality in Nonconvex-Nonconcave Minimax Optimization?
    Jin, Chi
    Netrapalli, Praneeth
    Jordan, Michael, I
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 119, 2020, 119
  • [7] Calm Local Optimality for Nonconvex-Nonconcave Minimax Problems
    Xiaoxiao Ma
    Wei Yao
    Jane J. Ye
    Jin Zhang
    Set-Valued and Variational Analysis, 2025, 33 (2)
  • [8] Proximal Point Methods and Nonconvex Optimization
    A. Kaplan
    R. Tichatschke
    Journal of Global Optimization, 1998, 13 : 389 - 406
  • [9] Proximal point methods and nonconvex optimization
    Kaplan, A
    Tichatschke, R
    JOURNAL OF GLOBAL OPTIMIZATION, 1998, 13 (04) : 389 - 406
  • [10] Limiting Behaviors of Nonconvex-Nonconcave Minimax Optimization via Continuous-Time Systems
    Grimmer, Benjamin
    Lu, Haihao
    Worah, Pratik
    Mirrokni, Vahab
    INTERNATIONAL CONFERENCE ON ALGORITHMIC LEARNING THEORY, VOL 167, 2022, 167