Successive Convex Approximation Algorithms for Sparse Signal Estimation With Nonconvex Regularizations

被引:28
|
作者
Yang, Yang [1 ]
Pesavento, Marius [2 ]
Chatzinotas, Symeon [1 ]
Ottersten, Bjorn [1 ]
机构
[1] Univ Luxembourg, Interdisciplinary Ctr Secur Reliabil & Trust, L-1855 Luxembourg, Luxembourg
[2] Tech Univ Darmstadt, Commun Syst Grp, D-64283 Darmstadt, Germany
基金
欧盟地平线“2020”;
关键词
Big data; line search; majorization minimization; nonconvex regularization; successive convex approximation; CONVERGENCE ANALYSIS; SHRINKAGE; SELECTION; FAMILY;
D O I
10.1109/JSTSP.2018.2877584
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
In this paper, we propose a successive convex approximation framework for sparse optimization where the nonsmooth regularization function in the objective function is nonconvex and it can be written as the difference of two convex functions. The proposed framework is based on a nontrivial combination of the majorization-minimization framework and the successive convex approximation framework proposed in literature for a convex regularization function. The proposed framework has several attractive features, namely, first, flexibility, as different choices of the approximate function lead to different types of algorithms; second, fast convergence, as the problem structure can be better exploited by a proper choice of the approximate function and the stepsize is calculated by the line search; third, low complexity, as the approximate function is convex and the line search scheme is carried out over a differentiable function; fourth, guaranteed convergence to a stationary point. We demonstrate these features by two example applications in subspace learning, namely the network anomaly detection problem and the sparse subspace clustering problem. Customizing the proposed framework by adopting the best-response type approximation, we obtain soft-thresholding with exact line search algorithms for which all elements of the unknown parameter are updated in parallel according to closed-form expressions. The attractive features of the proposed algorithms are illustrated numerically.
引用
收藏
页码:1286 / 1302
页数:17
相关论文
共 50 条
  • [41] LEARNING APPLIED TO SUCCESSIVE APPROXIMATION ALGORITHMS
    SARIDIS, GN
    IEEE TRANSACTIONS ON SYSTEMS SCIENCE AND CYBERNETICS, 1970, SSC6 (02): : 97 - &
  • [42] Sparse Approximation by Greedy Algorithms
    Temlyakov, V.
    MATHEMATICAL ANALYSIS, PROBABILITY AND APPLICATIONS - PLENARY LECTURES, 2016, 177 : 183 - 215
  • [43] SUCCESSIVE APPROXIMATION IN PARALLEL GRAPH ALGORITHMS
    FUSSELL, D
    THURIMELLA, R
    LECTURE NOTES IN COMPUTER SCIENCE, 1989, 349 : 205 - 217
  • [44] Novel successive-approximation algorithms
    Lampinen, H
    Perälä, P
    Vainio, O
    2005 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS), VOLS 1-6, CONFERENCE PROCEEDINGS, 2005, : 188 - 191
  • [45] SUCCESSIVE APPROXIMATION IN PARALLEL GRAPH ALGORITHMS
    FUSSELL, D
    THURIMELLA, R
    THEORETICAL COMPUTER SCIENCE, 1990, 74 (01) : 19 - 35
  • [46] Discretization and localization in successive convex relaxation methods for nonconvex quadratic optimization
    Kojima, M
    Tunçel, L
    MATHEMATICAL PROGRAMMING, 2000, 89 (01) : 79 - 111
  • [47] Sparse optimization for nonconvex group penalized estimation
    Lee, Sangin
    Oh, Miae
    Kim, Yongdai
    JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION, 2016, 86 (03) : 597 - 610
  • [48] Discretization and localization in successive convex relaxation methods for nonconvex quadratic optimization
    Masakazu Kojima
    Levent Tunçel
    Mathematical Programming, 2000, 89 : 79 - 111
  • [49] Nonconvex Sparse Regularization and Convex Optimization for Bearing Fault Diagnosis
    Wang, Shibin
    Selesnick, Ivan
    Cai, Gaigai
    Feng, Yining
    Sui, Xin
    Chen, Xuefeng
    IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, 2018, 65 (09) : 7332 - 7342
  • [50] SPARSE SIGNAL RECOVERY BASED ON NONCONVEX ENTROPY MINIMIZATION
    Huang, Shuai
    Tran, Dung N.
    Tran, Trac D.
    2016 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2016, : 3867 - 3871