CASIA OpenIR
Re-KISSME: A robust resampling scheme for distance metric learning in the presence of label noise
Zeng, Fanxia1,3; Zhang, Wensheng2,3; Zhang, Siheng2,3; Zheng, Nan1
Source PublicationNEUROCOMPUTING
ISSN0925-2312
2019-02-22
Volume330Pages:138-150
Corresponding AuthorZhang, Wensheng(wensheng.zhang@ia.ac.cn)
AbstractDistance metric learning aims to learn a metric with the similarity of samples. However, the increasing scalability and complexity of dataset or complex application brings about inevitable label noise, which frustrates the distance metric learning. In this paper, we propose a resampling scheme robust to label noise, Re-KISSME, based on Keep It Simple and Straightforward Metric (KISSME) learning method. Specifically, we consider the data structure and the priors of labels as two resampling factors to correct the observed distribution. By introducing the true similarity as latent variable, these two factors are integrated into a maximum likelihood estimation model. As a result, Re-KISSME can reason the underlying similarity of each pair and reduce the influence of label noise to estimate the metric matrix. Our model is solved by iterative algorithm with low computational cost. With synthetic label noise, the experiments on UCI datasets and two application datasets of person re-identification confirm the effectiveness of our proposal. (C) 2018 Elsevier B.V. All rights reserved.
KeywordResampling scheme KISSME Distance metric learning Label noise
DOI10.1016/j.neucom.2018.11.009
WOS KeywordPERSON REIDENTIFICATION ; CLASSIFICATION
Indexed BySCI
Language英语
Funding ProjectNational Key R&D Program of China[2017YFC0803700] ; National Natural Science Foundation of China[U1636220] ; National Natural Science Foundation of China[61602482] ; National Natural Science Foundation of China[61501463] ; Beijing Natural Science Foundation[4172063]
Funding OrganizationNational Key R&D Program of China ; National Natural Science Foundation of China ; Beijing Natural Science Foundation
WOS Research AreaComputer Science
WOS SubjectComputer Science, Artificial Intelligence
WOS IDWOS:000454789500014
PublisherELSEVIER SCIENCE BV
Citation statistics
Cited Times:2[WOS]   [WOS Record]     [Related Records in WOS]
Document Type期刊论文
Identifierhttp://ir.ia.ac.cn/handle/173211/25632
Collection中国科学院自动化研究所
Corresponding AuthorZhang, Wensheng
Affiliation1.Chinese Acad Sci, Inst Automat, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China
2.Chinese Acad Sci, Inst Automat, Res Ctr Precis Sensing & Control, Beijing 100190, Peoples R China
3.Univ Chinese Acad Sci, Beijing, Peoples R China
First Author AffilicationChinese Acad Sci, Inst Automat, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China
Corresponding Author AffilicationChinese Acad Sci, Inst Automat, Res Ctr Precis Sensing & Control, Beijing 100190, Peoples R China
Recommended Citation
GB/T 7714
Zeng, Fanxia,Zhang, Wensheng,Zhang, Siheng,et al. Re-KISSME: A robust resampling scheme for distance metric learning in the presence of label noise[J]. NEUROCOMPUTING,2019,330:138-150.
APA Zeng, Fanxia,Zhang, Wensheng,Zhang, Siheng,&Zheng, Nan.(2019).Re-KISSME: A robust resampling scheme for distance metric learning in the presence of label noise.NEUROCOMPUTING,330,138-150.
MLA Zeng, Fanxia,et al."Re-KISSME: A robust resampling scheme for distance metric learning in the presence of label noise".NEUROCOMPUTING 330(2019):138-150.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Zeng, Fanxia]'s Articles
[Zhang, Wensheng]'s Articles
[Zhang, Siheng]'s Articles
Baidu academic
Similar articles in Baidu academic
[Zeng, Fanxia]'s Articles
[Zhang, Wensheng]'s Articles
[Zhang, Siheng]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Zeng, Fanxia]'s Articles
[Zhang, Wensheng]'s Articles
[Zhang, Siheng]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.