sparse estimation methods are aimed at using or obtaining parsimonious
representations of data or models. While naturally cast as a combinatorial
optimization problem, variable or feature selection admits a convex relaxation
through the regularization by the $\ell_1$-norm. In this pape