Vector minimax concave penalty for sparse representation

Shibin Wang, Xuefeng Chen, Weiwei Dai, Ivan Selesnick, Gaigai Cai, Benjamin Cowen

Research output: Contribution to journalArticle

Abstract

This paper proposes vector minimax concave (VMC) penalty for sparse representation using tools of Moreau envelope. The VMC penalty is a weighted MC function; by fine tuning the weight of the VMC penalty with given strategy, the VMC regularized least squares problem shares the same global minimizers with the L0 regularization problem but has fewer local minima. Facilitated by the alternating direction method of multipliers (ADMM), the VMC regularization problem can be tackled as a sequence of convex sub-problems, each of which can be solved fast. Theoretical analysis of ADMM shows that the convergence of solving the VMC regularization problem is guaranteed. We present a series of numerical experiments demonstrating the superior performance of the VMC penalty and the ADMM algorithm in broad applications for sparse representation, including sparse denoising, sparse deconvolution, and missing data estimation.

Original languageEnglish (US)
Pages (from-to)165-179
Number of pages15
JournalDigital Signal Processing: A Review Journal
Volume83
DOIs
StatePublished - Dec 1 2018

Fingerprint

Deconvolution
Tuning
Experiments

Keywords

  • CEL0
  • Minimax-concave penalty
  • Nonconvex sparsity regularization
  • Sparse representation

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering

Cite this

Vector minimax concave penalty for sparse representation. / Wang, Shibin; Chen, Xuefeng; Dai, Weiwei; Selesnick, Ivan; Cai, Gaigai; Cowen, Benjamin.

In: Digital Signal Processing: A Review Journal, Vol. 83, 01.12.2018, p. 165-179.

Research output: Contribution to journalArticle

Wang, Shibin ; Chen, Xuefeng ; Dai, Weiwei ; Selesnick, Ivan ; Cai, Gaigai ; Cowen, Benjamin. / Vector minimax concave penalty for sparse representation. In: Digital Signal Processing: A Review Journal. 2018 ; Vol. 83. pp. 165-179.
@article{d5526c42e9344677b55028cfc315c9be,
title = "Vector minimax concave penalty for sparse representation",
abstract = "This paper proposes vector minimax concave (VMC) penalty for sparse representation using tools of Moreau envelope. The VMC penalty is a weighted MC function; by fine tuning the weight of the VMC penalty with given strategy, the VMC regularized least squares problem shares the same global minimizers with the L0 regularization problem but has fewer local minima. Facilitated by the alternating direction method of multipliers (ADMM), the VMC regularization problem can be tackled as a sequence of convex sub-problems, each of which can be solved fast. Theoretical analysis of ADMM shows that the convergence of solving the VMC regularization problem is guaranteed. We present a series of numerical experiments demonstrating the superior performance of the VMC penalty and the ADMM algorithm in broad applications for sparse representation, including sparse denoising, sparse deconvolution, and missing data estimation.",
keywords = "CEL0, Minimax-concave penalty, Nonconvex sparsity regularization, Sparse representation",
author = "Shibin Wang and Xuefeng Chen and Weiwei Dai and Ivan Selesnick and Gaigai Cai and Benjamin Cowen",
year = "2018",
month = "12",
day = "1",
doi = "10.1016/j.dsp.2018.08.021",
language = "English (US)",
volume = "83",
pages = "165--179",
journal = "Digital Signal Processing: A Review Journal",
issn = "1051-2004",
publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Vector minimax concave penalty for sparse representation

AU - Wang, Shibin

AU - Chen, Xuefeng

AU - Dai, Weiwei

AU - Selesnick, Ivan

AU - Cai, Gaigai

AU - Cowen, Benjamin

PY - 2018/12/1

Y1 - 2018/12/1

N2 - This paper proposes vector minimax concave (VMC) penalty for sparse representation using tools of Moreau envelope. The VMC penalty is a weighted MC function; by fine tuning the weight of the VMC penalty with given strategy, the VMC regularized least squares problem shares the same global minimizers with the L0 regularization problem but has fewer local minima. Facilitated by the alternating direction method of multipliers (ADMM), the VMC regularization problem can be tackled as a sequence of convex sub-problems, each of which can be solved fast. Theoretical analysis of ADMM shows that the convergence of solving the VMC regularization problem is guaranteed. We present a series of numerical experiments demonstrating the superior performance of the VMC penalty and the ADMM algorithm in broad applications for sparse representation, including sparse denoising, sparse deconvolution, and missing data estimation.

AB - This paper proposes vector minimax concave (VMC) penalty for sparse representation using tools of Moreau envelope. The VMC penalty is a weighted MC function; by fine tuning the weight of the VMC penalty with given strategy, the VMC regularized least squares problem shares the same global minimizers with the L0 regularization problem but has fewer local minima. Facilitated by the alternating direction method of multipliers (ADMM), the VMC regularization problem can be tackled as a sequence of convex sub-problems, each of which can be solved fast. Theoretical analysis of ADMM shows that the convergence of solving the VMC regularization problem is guaranteed. We present a series of numerical experiments demonstrating the superior performance of the VMC penalty and the ADMM algorithm in broad applications for sparse representation, including sparse denoising, sparse deconvolution, and missing data estimation.

KW - CEL0

KW - Minimax-concave penalty

KW - Nonconvex sparsity regularization

KW - Sparse representation

UR - http://www.scopus.com/inward/record.url?scp=85053290402&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053290402&partnerID=8YFLogxK

U2 - 10.1016/j.dsp.2018.08.021

DO - 10.1016/j.dsp.2018.08.021

M3 - Article

VL - 83

SP - 165

EP - 179

JO - Digital Signal Processing: A Review Journal

JF - Digital Signal Processing: A Review Journal

SN - 1051-2004

ER -