Accelerating ISTA with an active set strategy - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2011

Accelerating ISTA with an active set strategy

Résumé

Starting from a practical implementation of Roth and Fisher's algorithm to solve a Lasso-type problem, we propose and study the Active Set Iterative Shrinkage/Thresholding Algorithm (AS-ISTA). The convergence is proven by observing that the algorithm can be seen as a particular case of a coordinate gradient descent algorithm with a Gauss-Southwell-r rule. We provide experimental evidence that the proposed method can outperform FISTA and significantly speed-up the resolution of very undetermined inverse problems when using sparse convex priors. The proposed algorithm makes brain mapping with magneto- and electroencephalography (M/EEG) significantly faster when promoting spatially sparse and temporally smooth solutions using a ''group-Lasso'' $\ell_{21}$ mixed-norm.
Fichier principal
Vignette du fichier
KWGA_opt11.pdf (190.95 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00696992 , version 1 (14-05-2012)
hal-00696992 , version 2 (14-06-2013)

Identifiants

  • HAL Id : hal-00696992 , version 2

Citer

Matthieu Kowalski, Pierre Weiss, Alexandre Gramfort, Sandrine Anthoine. Accelerating ISTA with an active set strategy. OPT 2011: 4th International Workshop on Optimization for Machine Learning, Dec 2011, Sierra Nevada, Spain. pp.7. ⟨hal-00696992v2⟩
572 Consultations
221 Téléchargements

Partager

Gmail Facebook X LinkedIn More