On Sparsity Inducing Regularization Methods for Machine Learning