DocumentCode :
74222
Title :
Adaptive Randomized Coordinate Descent for Sparse Systems: Lasso and Greedy Algorithms
Author :
Onose, Alexandru ; Dumitrescu, Bogdan
Author_Institution :
Dept. of Signal Process., Tampere Univ. of Technol., Tampere, Finland
Volume :
63
Issue :
15
fYear :
2015
fDate :
Aug.1, 2015
Firstpage :
4091
Lastpage :
4101
Abstract :
Coordinate descent (CD) is a simple optimization technique suited to low complexity requirements and also for solving large problems. In randomized version, CD was recently shown as very effective for solving least-squares (LS) and other optimization problems. We propose here an adaptive version of randomized coordinate descent (RCD) for finding sparse LS solutions, from which we derive two algorithms, one based on the lasso criterion, the other using a greedy technique. Both algorithms employ a novel way of adapting the probabilities for choosing the coordinates, based on a matching pursuit criterion. Another new feature is that, in the lasso algorithm, the penalty term values are built without knowing the noise level or using other prior information. The proposed algorithms use efficient computations and have a tunable trade-off between complexity and performance through the number of CD steps per time instant. Besides a general theoretical convergence analysis, we present simulations that show good practical behavior, comparable to or better than that of state of the art methods.
Keywords :
greedy algorithms; iterative methods; least squares approximations; optimisation; probability; signal processing; adaptive RCD approach; adaptive randomized coordinate descent approach; general theoretical convergence analysis; greedy algorithm; lasso algorithm; least-square problem; low-complexity requirements; matching pursuit criterion; noise level; optimization technique; probabilities; signal processing application; sparse LS solution; sparse systems; Algorithm design and analysis; Approximation algorithms; Convergence; Greedy algorithms; Matching pursuit algorithms; Optimization; Signal processing algorithms; Adaptive algorithms; coordinate descent; randomization; sparse filters;
fLanguage :
English
Journal_Title :
Signal Processing, IEEE Transactions on
Publisher :
ieee
ISSN :
1053-587X
Type :
jour
DOI :
10.1109/TSP.2015.2436369
Filename :
7111352
Link To Document :
بازگشت