Asymptotic properties of dual averaging algorithm for constrained distributed stochastic optimization

被引:1
|
作者
Zhao, Shengchao [1 ]
Chen, Xing-Min [1 ]
Liu, Yongchao [1 ]
机构
[1] Dalian Univ Technol, Sch Math Sci, Dalian 116024, Peoples R China
基金
中国国家自然科学基金;
关键词
Constrained distributed stochastic optimization; Distributed dual averaging method; Almost sure convergence; Asymptotic normality; Asymptotic efficiency; MULTIAGENT OPTIMIZATION; RESOURCE-ALLOCATION; RANDOM NETWORKS; APPROXIMATION; CONVERGENCE;
D O I
10.1016/j.sysconle.2022.105252
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Considering the constrained stochastic optimization problem over a time-varying random network, where the agents are to collectively minimize a sum of objective functions subject to a common constraint set, we investigate asymptotic properties of a distributed algorithm based on dual averaging of gradients. Different from most existing works on distributed dual averaging algorithms that are mainly focused on their non-asymptotic properties, we prove not only almost sure convergence and rate of almost sure convergence, but also asymptotic normality and asymptotic efficiency of the algorithm. Firstly, for general constrained convex optimization problem distributed over a random network, we prove that almost sure consensus can be achieved and the estimates of agents converge to the same optimal point. For the case of linear constrained convex optimization, we show that the mirror map of the averaged dual sequence identifies the active constraints of the optimal solution with probability 1, which helps us to prove the almost sure convergence rate and then establish asymptotic normality of the algorithm. Furthermore, we also verify that the algorithm is asymptotically optimal. To the best of our knowledge, it is the first asymptotic normality result for constrained distributed optimization algorithms. Finally, a numerical example is provided to justify the theoretical analysis. (C) 2022 Elsevier B.V. All rights reserved.
引用
收藏
页数:14
相关论文
共 50 条
  • [1] ASYMPTOTIC PROPERTIES OF PRIMAL-DUAL ALGORITHM FOR DISTRIBUTED STOCHASTIC OPTIMIZATION OVER RANDOM NETWORKS WITH IMPERFECT COMMUNICATIONS
    Lei, Jinlong
    Chen, Han-Fu
    Fang, Hai-Tao
    SIAM JOURNAL ON CONTROL AND OPTIMIZATION, 2018, 56 (03) : 2159 - 2188
  • [2] A Primal-Dual Algorithm for Distributed Stochastic Optimization with Equality Constraints
    Du, Kai-Xin
    Chen, Xing-Min
    2021 PROCEEDINGS OF THE 40TH CHINESE CONTROL CONFERENCE (CCC), 2021, : 5586 - 5591
  • [3] Distributed Stochastic Projection-Free Algorithm for Constrained Optimization
    Jiang, Xia
    Zeng, Xianlin
    Xie, Lihua
    Sun, Jian
    Chen, Jie
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2025, 70 (04) : 2479 - 2494
  • [4] A Stochastic Gradient-Based Projection Algorithm for Distributed Constrained Optimization
    Zhang, Keke
    Gao, Shanfu
    Chen, Yingjue
    Zheng, Zuqing
    Lu, Qingguo
    NEURAL INFORMATION PROCESSING, ICONIP 2023, PT I, 2024, 14447 : 356 - 367
  • [5] Dual Averaging Methods for Regularized Stochastic Learning and Online Optimization
    Xiao, Lin
    JOURNAL OF MACHINE LEARNING RESEARCH, 2010, 11 : 2543 - 2596
  • [6] Accelerated Dual Averaging Methods for Decentralized Constrained Optimization
    Liu, Changxin
    Shi, Yang
    Li, Huiping
    Du, Wenli
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2023, 68 (04) : 2125 - 2139
  • [7] Asymptotic properties of distributed social sampling algorithm
    Qian LIU
    Xingkang HE
    Haitao FANG
    ScienceChina(InformationSciences), 2020, 63 (01) : 152 - 166
  • [8] Asymptotic properties of distributed social sampling algorithm
    Qian Liu
    Xingkang He
    Haitao Fang
    Science China Information Sciences, 2020, 63
  • [9] Asymptotic properties of distributed social sampling algorithm
    Liu, Qian
    He, Xingkang
    Fang, Haitao
    SCIENCE CHINA-INFORMATION SCIENCES, 2020, 63 (01)
  • [10] An Adaptive Primal-Dual Subgradient Algorithm for Online Distributed Constrained Optimization
    Yuan, Deming
    Ho, Daniel W. C.
    Jiang, Guo-Ping
    IEEE TRANSACTIONS ON CYBERNETICS, 2018, 48 (11) : 3045 - 3055