A decentralized Nesterov gradient method for stochastic optimization over unbalanced directed networks

被引:6
作者
Hu, Jinhui [1 ]
Xia, Dawen [2 ]
Cheng, Huqiang [1 ]
Feng, Liping [3 ]
Ji, Lianghao [4 ]
Guo, Jing [1 ]
Li, Huaqing [1 ]
机构
[1] Southwest Univ, Chongqing Key Lab Nonlinear Circuits & Intelligen, Coll Elect & Informat Engn, 2 Tiansheng Rd, Chongqing, Peoples R China
[2] Guizhou Minzu Univ, Coll Data Sci & Informat Engn, Guiyang, Peoples R China
[3] Xinzhou Teachers Univ, Dept Comp Sci, Xinzhou, Shanxi, Peoples R China
[4] Chongqing Univ Posts & Telecommun, Chongqing Key Lab Computat Intelligence, Chongqing, Peoples R China
基金
中国国家自然科学基金;
关键词
decentralized optimization; unbalanced directed networks; stochastic gradients; machine learning; multi‐ agent systems; TRACKING CONTROL; CONSENSUS; CONVERGENCE; ALGORITHM; SYSTEMS;
D O I
10.1002/asjc.2483
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Decentralized stochastic gradient methods play significant roles in large-scale optimization that finds many practical applications in machine learning and coordinated control. This paper studies optimization problems over unbalanced directed networks, where the mutual goal of agents in the network is to optimize a global objective function expressed as a sum of local objective functions. Each agent using only local computation and communication in the networks is assumed to get access to a stochastic first-order oracle. In order to devise a noise-tolerant decentralized algorithm with accelerated linear convergence, a decentralized Nesterov gradient algorithm with the constant step-size and parameter using stochastic gradients is proposed in this paper. The proposed algorithm employing a gradient-tracking technique is proved to converge linearly to an error ball around the optimal solution via the analysis on a linear system when the positive constant step-size and parameter are sufficiently small. We further recover the exact linear convergence for the proposed algorithm with exact gradients under the same selection conditions of the constant step-size and parameter. Some real-world data sets are used in simulations to validate the correctness of the theoretical findings and practicability of the proposed algorithm.
引用
收藏
页码:576 / 593
页数:18
相关论文
共 39 条
[1]   Convex Optimization: Algorithms and Complexity [J].
不详 .
FOUNDATIONS AND TRENDS IN MACHINE LEARNING, 2015, 8 (3-4) :232-+
[2]   Optimization Methods for Large-Scale Machine Learning [J].
Bottou, Leon ;
Curtis, Frank E. ;
Nocedal, Jorge .
SIAM REVIEW, 2018, 60 (02) :223-311
[3]  
Boyd S., 2011, DISTRIBUTED OPTIMIZA, DOI DOI 10.1561/2200000016
[4]   Coordinated Tracking Control With Asynchronous Edge-Based Event-Triggered Communications [J].
Cheng, Bin ;
Li, Zhongkui .
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2019, 64 (10) :4321-4328
[5]   On the convergence of exact distributed generalisation and acceleration algorithm for convex optimisation [J].
Cheng, Huqiang ;
Li, Huaqing ;
Wang, Zheng .
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE, 2020, 51 (16) :3408-3424
[6]  
Dua D, 2019, UCI U CALIFORNIA IRV
[7]  
Horn R., 1985, JOHNSON
[8]   DISTRIBUTED CONSENSUS FILTER FOR A CLASS OF CONTINUOUS-TIME NONLINEAR STOCHASTIC SYSTEMS IN SENSOR NETWORKS [J].
Jenabzadeh, Ahmadreza ;
Safarinejadian, Behrouz ;
Mohammadnia, Foroogh .
ASIAN JOURNAL OF CONTROL, 2017, 19 (04) :1284-1294
[9]  
LI H, 2020, IEEE T EMERG TOP COM
[10]   Distributed Robust Algorithm for Economic Dispatch in Smart Grids Over General Unbalanced Directed Networks [J].
Li, Huaqing ;
Wang, Zheng ;
Chen, Guo ;
Dong, Zhao Yang .
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, 2020, 16 (07) :4322-4332