Direct l_(2, p)-Norm Learning for Feature Selection

In this paper , we propose a novel sparse learning based feature selection method that directly optimizes a large margin linear classification model’s sparsity with -norm ( ) subject to data-fitt ing constraints, rather than using the sparsity as a regularizat ion term. To solve the d irect sparsity opt imizat ion p rob lem that is non-s mooth and non-convex when , we prov ide an efficient iterat ive algorithm with p roved convergence by convert ing it to a convex and s mooth optimizat ion prob lem at every iterat ion step . The proposed algorithm has been evaluated based on publicly availab le datasets , and extensive comparison experiments have demonstrated that our algorithm could ach ieve featu re select ion performance compet it ive to state-o f-the-art algorithms.