Non-Negative Orthogonal Greedy Algorithms

被引:32
作者
Thanh Thi Nguyen [1 ,2 ]
Idier, Jerome [3 ]
Soussen, Charles [4 ]
Djermoune, El-Hadi [1 ,2 ]
机构
[1] Univ Lorraine, F-54506 Vandceuvre Les Nancy, France
[2] CNRS, Ctr Rech Automat Nancy, F-54506 Vandceuvre Les Nancy, France
[3] CNRS, Lab Sci Numer Nantes, F-44321 Nantes, France
[4] Univ Paris Saclay, Univ Paris Sud, CNRS, Cent Supelec,Lab Signaux & Syst, F-91192 Gif Sur Yvette, France
关键词
Greedy algorithms; Signal processing algorithms; Matching pursuit algorithms; Sparse representation; Approximation algorithms; Dictionaries; Acceleration; Orthogonal greedy algorithms; sparse reconstru-ction; non-negativity; non-negative least-squares; active-set algorithms; LEAST-SQUARES; SPARSE RECOVERY; MATRIX FACTORIZATION; PURSUIT;
D O I
10.1109/TSP.2019.2943225
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
Orthogonal greedy algorithms are popular sparse signal reconstruction algorithms. Their principle is to select atoms one by one. A series of unconstrained least-square subproblems of gradually increasing size is solved to compute the approximation coefficients, which is efficiently performed using a fast recursive update scheme. When dealing with non-negative sparse signal reconstruction, a series of non-negative least-squares subproblems have to be solved. Fast implementation becomes tricky since each subproblem does not have a closed-form solution anymore. Recently, non-negative extensions of the classical orthogonal matching pursuit and orthogonal least squares algorithms were proposed, using slow (i.e., non-recursive) or recursive but inexact implementations. In this paper, we revisit these algorithms in a unified way. We define a class of non-negative orthogonal greedy algorithms and exhibit their structural properties. We propose a fast and exact implementation based on the active-set resolution of non-negative least-squares and exploiting warm start initializations. The algorithms are assessed in terms of accuracy and computational complexity for a sparse spike deconvolution problem. We also present an application to near-infrared spectra decomposition.
引用
收藏
页码:5643 / 5658
页数:16
相关论文
共 52 条
[1]  
[Anonymous], FOUND TRENDS MACH LE
[2]  
[Anonymous], 2013, RECENT ADV HARMONIC
[3]  
[Anonymous], 2002, SUBSET SELECTION REG
[4]  
Belmerhnia L, 2015, 2015 IEEE 6TH INTERNATIONAL WORKSHOP ON COMPUTATIONAL ADVANCES IN MULTI-SENSOR ADAPTIVE PROCESSING (CAMSAP), P437, DOI 10.1109/CAMSAP.2015.7383830
[5]   Nonnegative least-squares image deblurring: improved gradient projection approaches [J].
Benvenuto, F. ;
Zanella, R. ;
Zanni, L. ;
Bertero, M. .
INVERSE PROBLEMS, 2010, 26 (02)
[6]  
Bjorck A, 1996, NUMERICAL METHODS LE, DOI DOI 10.1137/1.9781611971484
[7]  
Blumensath Thomas, 2007, TECH REP
[8]   Dynamic Screening: Accelerating First-Order Algorithms for the Lasso and Group-Lasso [J].
Bonnefoy, Antoine ;
Emiya, Valentin ;
Ralaivola, Liva ;
Gribonval, Remi .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2015, 63 (19) :5121-5132
[9]  
Bro R, 1997, J CHEMOMETR, V11, P393, DOI 10.1002/(SICI)1099-128X(199709/10)11:5<393::AID-CEM483>3.0.CO
[10]  
2-L