文献类型：期刊文献

英文题名：Hierarchical sparse representation based Multi-Instance Semi-Supervised Learning with application to image categorization

作者：Feng, Songhe[1,2];Xiong, Weihua[3];Li, Bing[3];Lang, Congyan[1];Huang, Xiankai[4]

第一作者：Feng, Songhe

通讯作者：Feng, SH[1]

机构：[1]Beijing Jiaotong Univ, Sch Comp & Informat Technol, Beijing, Peoples R China;[2]Beijing Jiaotong Univ, Beijing Key Lab Transportat Data Anal & Min, Beijing, Peoples R China;[3]Chinese Acad Sci, NLPR, Inst Automat, Beijing, Peoples R China;[4]Beijing Union Univ, Tourism Inst, Beijing, Peoples R China

第一机构：Beijing Jiaotong Univ, Sch Comp & Informat Technol, Beijing, Peoples R China

通讯机构：[1]corresponding author), Beijing Jiaotong Univ, Sch Comp & Informat Technol, Beijing, Peoples R China.

年份：2014

卷号：94

期号：1

起止页码：595-607

外文期刊名：SIGNAL PROCESSING

收录：;EI(收录号：20133716729074);Scopus(收录号：2-s2.0-84883173836);WOS:【SCI-EXPANDED(收录号:WOS:000327363300062)】；

基金：This work is supported by National Nature Science Foundation of China (61100142, 61272352, 61271369, 61372148, 61370038), the Fundamental Research Funds for the Central Universities (2011JBM218, 2012JBM040), the Doctoral Fund of Ministry of Education of China (20110009120005) and the Science Foundation of Beijing Jiaotong University (2012RC008).

语种：英文

外文关键词：Multi-Instance Semi-Supervised Learning; Hierarchical sparse representation; Weighted multi-instance kernel; Image categorization

摘要：Recent studies have shown that sparse representation (SR) can deal well with many computer vision problems. In this paper, we extend a hierarchical sparse representation algorithm into Multi-Instance Semi-Supervised Learning (MISSL) problem. Specifically, at the instance level, after investigating the properties of true positive instances in depth, we propose a novel instance disambiguation strategy based on sparse representation that can identify the instance confidence value in both positive and unlabeled bags more effectively. At the bag level, in contrast to the traditional k-NN or epsilon-graph construction methods used in the graph-based semi-supervised learning settings, we propose a weighted multi-instance kernel and a corresponding kernel sparse representation method for robust l(1)-graph construction. The improved e(1)-graph that encodes the multi-instance properties can be utilized in the manifold regularization framework for the label propagation. Experimental results on different image data sets have demonstrated that the proposed algorithm outperforms existing multi-instance learning (MIL) algorithms, as well as the MISSL algorithms with the application to image categorization task. (C) 2013 Elsevier B.V. All rights reserved.