<i>l</i>₀-norm based structural sparse least square regression for feature selection

作     者：Han, Jiuqi Sun, Zhengya Hao, Hongwei 

作者机构：Chinese Acad Sci Inst Automat 95 Zhongguancun East Rd Beijing 100190 Peoples R China 

出 版 物：《PATTERN RECOGNITION》 (图形识别)

年 卷 期：2015年第48卷第12期

页      面：3927-3940页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China Hundred Talents Program (Chinese Academy of Sciences) [Y3S4011D31] 

主　　题：Structural sparse learning l(o)-norm Least square regression Feature selection Adaptive greedy algorithm 

摘      要：This paper presents a novel approach for feature selection with regard to the problem of structural sparse least square regression (SSLSR). Rather than employing the l(1)-norm regularization to control the sparsity, we directly work with sparse solutions via l(o)-norm regularization. In particular, we develop an effective greedy algorithm, where the forward and backward steps are combined adaptively, to resolve the SSLSR problem with the intractable l(r,o)-norm. On the one hand, features with the strongest correlation to classes are selected in the forward steps. On the other hand, redundant features which contribute little to the improvement of the objective function are removed in the backward steps. Furthermore, we provide solid theoretical analysis to prove the effectiveness of the proposed method. Experimental results on synthetic and real world data sets from different domains also demonstrate the superiority of the proposed method over the state-of-the-arts. (C) 2015 Elsevier Ltd. All rights reserved.