Efficient Sparse Representation and Modeling

Abstract

In this chapter, some efficient sparse representation algorithms including feature-sign algorithm, graphical modeling methods and efficient sparse Bayesian learning, sparse quantization, hashed sparse representation, and compressive feature methods are described. Feature-sign search method uses the greedy algorithm’s way to solve the constraint model. The graphical model can be used to improve the speed of the algorithm and it can be efficiently used in Bayesian compressed sensing. Efficient sparse Bayesian Learning also uses the greedy algorithm way to solve sparse Bayesian learning model. Sparse quantization can be used efficiently to quantify the features which can be used in classification and can save the memory of the computer. Hashed sparse representation uses hashing way to efficiently search the nonzero position of the sparse representation. Compressive feature uses the theory of compressed sensing to compress the feature which can reduce the computational cost of the algorithms.