On the Convergence of an Alternating Direction Penalty Method for Nonconvex Problems

被引:0
|
作者
Magnusson, S. [1 ]
Weeraddana, P. C. [1 ]
Rabbat, M. G. [2 ]
Fischione, C. [1 ]
机构
[1] KTH Royal Inst Technol, Dept Automat Control, Stockholm, Sweden
[2] McGill Univ, Dept Elect & Comp Engn, Montreal, PQ, Canada
来源
CONFERENCE RECORD OF THE 2014 FORTY-EIGHTH ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS & COMPUTERS | 2014年
关键词
Nonconvex Optimization; Distributed Optimization; ALGORITHM; CONSENSUS;
D O I
暂无
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
This paper investigates convergence properties of scalable algorithms for nonconvex and structured optimization. We consider a method that is adapted from the classic quadratic penalty function method, the Alternating Direction Penalty Method (ADPM). Unlike the original quadratic penalty function method, in which single-step optimizations are adopted, ADPM uses alternating optimization, which in turn is exploited to enable scalability of the algorithm. A special case of ADPM is a variant of the well known Alternating Direction Method of Multipliers (ADMM), where the penalty parameter is increased to infinity. We show that due to the increasing penalty, the ADPM asymptotically reaches a primal feasible point under mild conditions. Moreover, we give numerical evidence that demonstrates the potential of the ADPM for computing local optimal points when the penalty is not updated too aggressively.
引用
收藏
页码:793 / 797
页数:5
相关论文
共 50 条
  • [1] On the Convergence of Alternating Direction Lagrangian Methods for Nonconvex Structured Optimization Problems
    Magnusson, Sindri
    Weeraddana, Pradeep Chathuranga
    Rabbat, Michael G.
    Fischione, Carlo
    IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, 2016, 3 (03): : 296 - 309
  • [2] CONVERGENCE ANALYSIS OF ALTERNATING DIRECTION METHOD OF MULTIPLIERS FOR A FAMILY OF NONCONVEX PROBLEMS
    Hong, Mingyi
    Luo, Zhi-Quan
    Razaviyayn, Meisam
    SIAM JOURNAL ON OPTIMIZATION, 2016, 26 (01) : 337 - 364
  • [3] Alternating direction method of multipliers for nonconvex fused regression problems
    Xiu, Xianchao
    Liu, Wanquan
    Li, Ling
    Kong, Lingchen
    COMPUTATIONAL STATISTICS & DATA ANALYSIS, 2019, 136 : 59 - 71
  • [4] A new alternating direction method for linearly constrained nonconvex optimization problems
    Wang, X. Y.
    Li, S. J.
    Kou, X. P.
    Zhang, Q. F.
    JOURNAL OF GLOBAL OPTIMIZATION, 2015, 62 (04) : 695 - 709
  • [5] A Symmetric Alternating Direction Method of Multipliers for Separable Nonconvex Minimization Problems
    Wu, Zhongming
    Li, Min
    Wang, David Z. W.
    Han, Deren
    ASIA-PACIFIC JOURNAL OF OPERATIONAL RESEARCH, 2017, 34 (06)
  • [6] Computing Feasible Points of Bilevel Problems with a Penalty Alternating Direction Method
    Kleinert, Thomas
    Schmidt, Martin
    INFORMS JOURNAL ON COMPUTING, 2021, 33 (01) : 198 - 215
  • [7] Convergence of the RMSProp deep learning method with penalty for nonconvex optimization
    Xu, Dongpo
    Zhang, Shengdong
    Zhang, Huisheng
    Mandic, Danilo P.
    NEURAL NETWORKS, 2021, 139 : 17 - 23
  • [8] Alternating direction method of multipliers for a class of nonconvex bilinear optimization: convergence analysis and applications
    Davood Hajinezhad
    Qingjiang Shi
    Journal of Global Optimization, 2018, 70 : 261 - 288
  • [9] Convergence of alternating direction method for minimizing sum of two nonconvex functions with linear constraints
    Guo, K.
    Han, D. R.
    Wu, T. T.
    INTERNATIONAL JOURNAL OF COMPUTER MATHEMATICS, 2017, 94 (08) : 1653 - 1669
  • [10] Nonconvex Sparse Spectral Clustering by Alternating Direction Method of Multipliers and Its Convergence Analysis
    Lu, Canyi
    Feng, Jiashi
    Lin, Zhouchen
    Yan, Shuicheng
    THIRTY-SECOND AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTIETH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / EIGHTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2018, : 3714 - 3721