Paper Info
Title
Total Variation Minimization for Compressed Sensing with “Smoothly” Varying Covariates
Abstract
The LASSO is a variable subset selection procedure in statistical linear regression based on sparsity promoting `1 penalization of the least-squares operator. In many applications, the design matrix has strongly correlated columns which are smoothly evolving with the column index. For such applications, the standard LASSO does not provide satisfactory solutions in practice because some incoherence is often needed for support recovery of sparse vectors. In this paper, we circumvent this problem by using a Total Variation penalty and obtain adaptive confidence intervals for the nonzero components of the signal. The relaxation parameter is calibrated by a new multiscale data acquisition scheme. This approach is illustrated by some simulations results for a source localization problem in a marine environment.
Year
DOI
Venue
2016
10.1109/CSE-EUC-DCABES.2016.233
2016 IEEE Intl Conference on Computational Science and Engineering (CSE) and IEEE Intl Conference on Embedded and Ubiquitous Computing (EUC) and 15th Intl Symposium on Distributed Computing and Applications for Business Engineering (DCABES)
Keywords
Field
DocType
total variation minimization,compressed sensing,smoothly varying covariates,LASSO,variable subset selection,statistical linear regression,least-squares operator,design matrix,column index,support recovery,sparse vectors,total variation penalty,adaptive confidence intervals,nonzero components,relaxation parameter,multiscale data acquisition scheme,source localization problem,marine environment
Mathematical optimization,Covariate,Feature selection,Computer science,Lasso (statistics),Minification,Design matrix,Sparse matrix,Compressed sensing,Linear regression
Conference
ISBN
Citations 
PageRank 
978-1-5090-3594-6
0
0.34
References 
Authors
8
3
Name
Order
Citations
PageRank
Stéphane Chrétien11110.68
Peter Harris200.34
Rami Tawil3364.36