Noise-robust semi-supervised learning via fast sparse coding

doi:10.1016/j.patcog.2014.08.019

CORC > 北京大学 > 信息科学技术学院

	Noise-robust semi-supervised learning via fast sparse coding
	Lu, Zhiwu ; Wang, Liwei
刊名	模式识别
	2015
关键词	Graph-based semi-supervised learning Noise reduction Laplacian regularization Sparse coding Noise-robust image classification LABEL PROPAGATION INVERSE PROBLEMS REPRESENTATION SELECTION GRAPH REGULARIZATION SHRINKAGE ALGORITHM
DOI	10.1016/j.patcog.2014.08.019
英文摘要	This paper presents a novel noise-robust graph-based semi-supervised learning algorithm to deal with the challenging problem of semi-supervised learning with noisy initial labels. Inspired by the successful use of sparse coding for noise reduction, we choose to give new L-1-norm formulation of Laplacian regularization for graph-based semi-supervised learning. Since our L-1-norm Laplacian regularization is explicitly defined over the eigenvectors of the normalized Laplacian matrix, we formulate graph-based semi-supervised learning as an L-1-norm linear reconstruction problem which can be efficiently solved by sparse coding. Furthermore, by working with only a small subset of eigenvectors, we develop a fast sparse coding algorithm for our L-1-norm semi-supervised learning. Finally, we evaluate the proposed algorithm in noise-robust image classification. The experimental results on several benchmark datasets demonstrate the promising performance of the proposed algorithm. (C) 2014 Elsevier Ltd. All rights reserved.; http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000345809600028&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=8e1609b174ce4e31116a60747a720701 ; Computer Science, Artificial Intelligence; Engineering, Electrical & Electronic; SCI(E); 6; ARTICLE; zhiwu.lu@gmail.com; wanglw@cis.pku.edu.cn; 2; 605-612; 48
语种	英语
内容类型	期刊论文
源URL	[http://ir.pku.edu.cn/handle/20.500.11897/206235]
专题	信息科学技术学院
推荐引用方式 GB/T 7714	Lu, Zhiwu,Wang, Liwei. Noise-robust semi-supervised learning via fast sparse coding[J]. 模式识别,2015.
APA	Lu, Zhiwu,&Wang, Liwei.(2015).Noise-robust semi-supervised learning via fast sparse coding.模式识别.
MLA	Lu, Zhiwu,et al."Noise-robust semi-supervised learning via fast sparse coding".模式识别 (2015).