Bayesian and L$_\mathbf{1}$ Approaches to Sparse Unsupervised Learning