A Unified Analysis of Extra-gradient and Optimistic Gradient Methods for Saddle Point Problems: Proximal Point Approach

被引:0
|
作者
Mokhtari, Aryan [1 ]
Ozdaglar, Asuman [2 ]
Pattathil, Sarath [2 ]
机构
[1] UT Austin, Austin, TX 78712 USA
[2] MIT, Cambridge, MA 02139 USA
来源
INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 108 | 2020年 / 108卷
关键词
LINEAR CONVERGENCE; MONOTONE-OPERATORS; ALGORITHM;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper we consider solving saddle point problems using two variants of Gradient Descent-Ascent algorithms, Extra-gradient (EG) and Optimistic Gradient Descent Ascent (OGDA) methods. We show that both of these algorithms admit a unified analysis as approximations of the classical proximal point method for solving saddle point problems. This viewpoint enables us to develop a new framework for analyzing EG and OGDA for bilinear and strongly convex-strongly concave settings. Moreover, we use the proximal point approximation interpretation to generalize the results for OGDA for a wide range of parameters.
引用
收藏
页码:1497 / 1506
页数:10
相关论文
共 50 条
  • [1] Tight Analysis of Extra-gradient and Optimistic Gradient Methods For Nonconvex Minimax Problems
    Mahdavinia, Pouria
    Deng, Yuyang
    Li, Haochuan
    Mahdavi, Mehrdad
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [2] Extra-gradient methods for solving split feasibility and fixed point problems
    Chen, Jin-Zuo
    Ceng, Lu-Chuan
    Qiu, Yang-Qing
    Kong, Zhao-Rong
    FIXED POINT THEORY AND APPLICATIONS, 2015,
  • [3] Extra-gradient methods for solving split feasibility and fixed point problems
    Jin-Zuo Chen
    Lu-Chuan Ceng
    Yang-Qing Qiu
    Zhao-Rong Kong
    Fixed Point Theory and Applications, 2015
  • [4] Analysis of iterative methods for saddle point problems: A unified approach
    Zulehner, W
    MATHEMATICS OF COMPUTATION, 2002, 71 (238) : 479 - 505
  • [5] Proximal Regularization for the Saddle Point Gradient Dynamics
    Goldsztajn, Diego
    Paganini, Fernando
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2021, 66 (09) : 4385 - 4392
  • [6] Continuous Projection Generalized Extra-Gradient Quasi-Newton Second-Order Method for Solving Saddle Point Problems
    V. G. Malinov
    Computational Mathematics and Mathematical Physics, 2022, 62 : 753 - 765
  • [7] Continuous Projection Generalized Extra-Gradient Quasi-Newton Second-Order Method for Solving Saddle Point Problems
    Malinov, V. G.
    COMPUTATIONAL MATHEMATICS AND MATHEMATICAL PHYSICS, 2022, 62 (05) : 753 - 765
  • [8] Convergence analysis of accelerated proximal extra-gradient method with applications
    Verma, Mridula
    Shukla, K. K.
    NEUROCOMPUTING, 2020, 388 : 288 - 300
  • [9] CONVERGENCE RATE OF O(1/k) FOR OPTIMISTIC GRADIENT AND EXTRAGRADIENT METHODS IN SMOOTH CONVEX-CONCAVE SADDLE POINT PROBLEMS
    Mokhtari, Aryan
    Ozdaglar, Asuman E.
    Pattathil, Sarath
    SIAM JOURNAL ON OPTIMIZATION, 2020, 30 (04) : 3230 - 3251
  • [10] Subgradient Extra-Gradient Algorithm for Pseudomonotone Equilibrium Problems and Fixed-Point Problems of Bregman Relatively Nonexpansive Mappings
    Lotfikar, Roushanak
    Eskandani, Gholamreza Zamani
    Kim, Jong-Kyu
    Rassias, Michael Th.
    MATHEMATICS, 2023, 11 (23)