Sparse methods such as regularization by the L1-norm has attracted a lot of interest in recent years in statistics, machine learning and signal processing. In the context of least-square linear regres
