Unbiased Group-Sparsity Sensing Using Quadratic Envelopes - Archive ouverte HAL Access content directly
Conference Papers Year : 2019

Unbiased Group-Sparsity Sensing Using Quadratic Envelopes

Abstract

This paper investigates a new regularization of the group-sparsity estimation problem based on a quadratic envelope operator. The resulting estimator is shown to have a reduced bias when compared to the classical LASSO estimator and is characterized by a simple hyperparameter selection. Numerical results show that the quadratic envelope regularization yields estimates equal to an oracle solution with high probability. The robustness of the proposed hyperparameter selection rule is also analyzed.
Fichier principal
Vignette du fichier
CarlssonCAMSAP2019.pdf (620.3 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03034133 , version 1 (01-12-2020)

Identifiers

Cite

Marcus Carlsson, Jean-Yves Tourneret, Herwig Wendt. Unbiased Group-Sparsity Sensing Using Quadratic Envelopes. IEEE 8th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP 2019), Dec 2019, Le Gosier ;Guadeloupe, France. pp.425-429, ⟨10.1109/CAMSAP45676.2019.9022465⟩. ⟨hal-03034133⟩
10 View
84 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More