Proximal Stochastic Recursive Momentum Methods for Nonconvex Composite Decentralized Optimization

被引:0
作者
Mancino-Ball, Gabriel [1 ]
Miao, Shengnan [1 ]
Xu, Yangyang [1 ]
Chen, Jie [2 ]
机构
[1] Rensselaer Polytech Inst, Dept Math Sci, Troy, NY 12180 USA
[2] MIT, IBM Res, IBM Watson AI Lab, Cambridge, MA 02142 USA
来源
THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 7 | 2023年
关键词
CONVERGENCE;
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
Consider a network of N decentralized computing agents collaboratively solving a nonconvex stochastic composite problem. In this work, we propose a single-loop algorithm, called DEEPSTORM, that achieves optimal sample complexity for this setting. Unlike double-loop algorithms that require a large batch size to compute the (stochastic) gradient once in a while, DEEPSTORM uses a small batch size, creating advantages in occasions such as streaming data and online learning. requiring O(1) batch size. We conduct convergence analysis for DEEPSTORM with both constant and diminishing step sizes. Additionally, under proper initialization and a small enough desired solution error, we show that DEEPSTORM with a constant step size achieves a network-independent sample complexity, with an additional linear speed-up with respect to N over centralized methods. All codes are made available at https://github.com/gmancino/DEEPSTORM.
引用
收藏
页码:9055 / 9063
页数:9
相关论文
共 50 条
  • [31] Decentralized Stochastic Optimization With Random Attendance
    Tran Thi Phuong
    Le Trieu Phong
    [J]. IEEE SIGNAL PROCESSING LETTERS, 2022, 29 : 1322 - 1326
  • [32] Momentum methods for stochastic optimization over time-varying directed networks
    Cui, Zhuo-Xu
    Fan, Qibin
    Jia, Cui
    [J]. SIGNAL PROCESSING, 2020, 174
  • [33] Decentralized Inexact Proximal Gradient Method With Network-Independent Stepsizes for Convex Composite Optimization
    Guo, Luyao
    Shi, Xinli
    Cao, Jinde
    Wang, Zihao
    [J]. IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2023, 71 : 786 - 801
  • [34] Decentralized Dual Proximal Gradient Algorithms for Non-Smooth Constrained Composite Optimization Problems
    Li, Huaqing
    Hu, Jinhui
    Ran, Liang
    Wang, Zheng
    Lu, Qingguo
    Du, Zhenyuan
    Huang, Tingwen
    [J]. IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, 2021, 32 (10) : 2594 - 2605
  • [35] Stochastic subgradient projection methods for composite optimization with functional constraints
    Necoara, Ion
    Singh, Nitesh Kumar
    [J]. JOURNAL OF MACHINE LEARNING RESEARCH, 2022, 23
  • [36] Conditional gradient type methods for composite nonlinear and stochastic optimization
    Ghadimi, Saeed
    [J]. MATHEMATICAL PROGRAMMING, 2019, 173 (1-2) : 431 - 464
  • [37] PROXIMAL QUASI-NEWTON METHODS FOR THE COMPOSITE MULTIOBJECTIVE OPTIMIZATION PROBLEMS
    Peng, Jianwen
    Ren, Jie
    Yao, Jen-Chih
    [J]. JOURNAL OF NONLINEAR AND CONVEX ANALYSIS, 2024, 25 (01) : 207 - 221
  • [38] Composite Optimization by Nonconvex Majorization-Minimization
    Geiping, Jonas
    Moeller, Michael
    [J]. SIAM JOURNAL ON IMAGING SCIENCES, 2018, 11 (04): : 2494 - 2528
  • [39] Randomized Block Proximal Methods for Distributed Stochastic Big-Data Optimization
    Farina, Francesco
    Notarstefano, Giuseppe
    [J]. IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2021, 66 (09) : 4000 - 4014
  • [40] The Boosted Double-proximal Subgradient Algorithm for nonconvex optimization
    Aragon-Artacho, Francisco J.
    Perez-Aros, Pedro
    Torregrosa-Belen, David
    [J]. MATHEMATICAL PROGRAMMING, 2025,