The problem of learning a latent model for sparse or low-dimensional representation of high-dimensional data has attracted significant attention for many years. This thesis focuses on learning latent models for sparse or low-dimensional representation of images, dynamic data, and documents with...