Learning sparsity pattern in high dimension is a great challenge in both implementation and theory. In this thesis we develop scalable Bayesian algorithms based on EM algorithm and variational inference to learn sparsity ...