Coherence-Based Performance Guarantees for Estimating a Sparse Vector Under Random Noise

被引:178
作者
Ben-Haim, Zvika [1 ]
Eldar, Yonina C.
Elad, Michael [2 ]
机构
[1] Technion Israel Inst Technol, Dept Elect Engn, IL-32000 Haifa, Israel
[2] Technion Israel Inst Technol, Dept Comp Sci, IL-32000 Haifa, Israel
基金
以色列科学基金会;
关键词
Basis pursuit; Dantzig selector; matching pursuit; oracle; sparse estimation; thresholding algorithm; L(1) MINIMIZATION; UNCERTAINTY RELATIONS; DANTZIG SELECTOR; STABLE RECOVERY; SIGNAL RECOVERY; LASSO; REPRESENTATIONS; PURSUIT;
D O I
10.1109/TSP.2010.2052460
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
We consider the problem of estimating a deterministic sparse vector x(0) from underdetermined measurements Ax(0) + w, where w represents white Gaussian noise and A is a given deterministic dictionary. We provide theoretical performance guarantees for three sparse estimation algorithms: basis pursuit denoising (BPDN), orthogonal matching pursuit (OMP), and thresholding. The performance of these techniques is quantified as the l(2) distance between the estimate and the true value of x(0). We demonstrate that, with high probability, the analyzed algorithms come close to the behavior of the oracle estimator, which knows the locations of the nonzero elements in x(0). Our results are non-asymptotic and are based only on the coherence of A, so that they are applicable to arbitrary dictionaries. This provides insight on the advantages and drawbacks of l(1) relaxation techniques such as BPDN and the Dantzig selector, as opposed to greedy approaches such as OMP and thresholding.
引用
收藏
页码:5030 / 5043
页数:14
相关论文
共 37 条
[1]  
[Anonymous], 1979, Generalized inverses of linear transformations
[2]  
[Anonymous], 2007, Found. Trends in Sig. Pro, DOI [DOI 10.1561/2000000008, 10.1561/2000000008]
[3]   COHERENCE-BASED NEAR-ORACLE PERFORMANCE GUARANTEES FOR SPARSE ESTIMATION UNDER GAUSSIAN NOISE [J].
Ben-Haim, Zvika ;
Eldar, Yonina C. ;
Elad, Michael .
2010 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, 2010, :3590-3593
[4]   The Cramr-Rao Bound for Estimating a Sparse Parameter Vector [J].
Ben-Haim, Zvika ;
Eldar, Yonina C. .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2010, 58 (06) :3384-3389
[5]   SIMULTANEOUS ANALYSIS OF LASSO AND DANTZIG SELECTOR [J].
Bickel, Peter J. ;
Ritov, Ya'acov ;
Tsybakov, Alexandre B. .
ANNALS OF STATISTICS, 2009, 37 (04) :1705-1732
[6]   From Sparse Solutions of Systems of Equations to Sparse Modeling of Signals and Images [J].
Bruckstein, Alfred M. ;
Donoho, David L. ;
Elad, Michael .
SIAM REVIEW, 2009, 51 (01) :34-81
[7]   On Recovery of Sparse Signals Via l1 Minimization [J].
Cai, T. Tony ;
Xu, Guangwu ;
Zhang, Jun .
IEEE TRANSACTIONS ON INFORMATION THEORY, 2009, 55 (07) :3388-3397
[8]   Stable Recovery of Sparse Signals and an Oracle Inequality [J].
Cai, Tony Tony ;
Wang, Lie ;
Xu, Guangwu .
IEEE TRANSACTIONS ON INFORMATION THEORY, 2010, 56 (07) :3516-3522
[9]  
CANDES EJ, 2006, ACTA NUMER, P1
[10]  
Candes E, 2007, ANN STAT, V35, P2313, DOI 10.1214/009053606000001523