Online Learning for Matrix Factorization and Sparse Coding

摘要: Sparse coding--that is, modelling data vectors as sparse linear combinations of basis elements--is widely used in machine learning, neuroscience, signal processing, and statistics. This paper focuses on the large-scale matrix factorization problem that consists learning set order to adapt it specific data. Variations this include dictionary non-negative principal component analysis. In paper, we propose address these tasks with a new online optimization algorithm, based stochastic approximations, which scales up gracefully large sets millions training samples, extends naturally various formulations, making suitable for wide range problems. A proof convergence is presented, along experiments natural images genomic demonstrating leads state-of-the-art performance terms speed both small sets.