L1-Norm Sparse Learning and Its Application

The need on finding sparse representations has attracted more and more people to research it. Researchers have developed many approaches (such as nonnegative constraint, 11-norm sparsity regularization and sparse Bayesian learning with independent Gaussian prior) for encouraging sparse solutions and established some conditions under which the feasible solutions could be found by those approaches. This paper commbined the L1-norm regularization and bayesian learning, called L1-norm sparse bayesian learning, which was inspired by RVM (relative vector machine). L1-norm sparse bayesian learning has found its applications in many fields such as MCR (multivariate curve resolution) and so on. We proposed a new method called BSMCR (bayesian sparse MCR) to enhance the quality of resolve result.