Learning kernel subspace classifier

Bailing Zhang, Hanseok Ko, Yongsheng Gao

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Subspace classifiers are well-known in pattern recognition, which represent pattern classes by linear subspaces spanned by the class specific basis vectors through simple mathematical operations like SVD. Recently, kernel based subspace methods have been proposed to extend the functionalities by directly applying the Kernel Principal Component Analysis (KPCA). The projection variance in kernel space as applied in these earlier proposed kernel subspace methods, however, is not a trustworthy criteria for class discrimination and they simply fail in many recognition problems as we encountered in biometrics research. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in input space through the kernel subspace projection. While the pre-image methods aiming at finding an approximate pre-image for each input by minimization of the reconstruction error in kernel space, we emphasize the problem of how to estimate a kernel subspace as a model for a specific class. Using the occluded face recognition as examples, our experimental results demonstrated the efficiency of the proposed method.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages299-308
Number of pages10
Volume4642 LNCS
Publication statusPublished - 2007 Dec 1
Event2007 International Conference on Advances in Biometrics, ICB 2007 - Seoul, Korea, Republic of
Duration: 2007 Aug 272007 Aug 29

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4642 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other2007 International Conference on Advances in Biometrics, ICB 2007
CountryKorea, Republic of
CitySeoul
Period07/8/2707/8/29

Fingerprint

Classifiers
Classifier
Subspace
Learning
kernel
Singular value decomposition
Biometrics
Face recognition
Principal component analysis
Pattern recognition
Subspace Methods
Principal Component Analysis
Projection
Kernel Principal Component Analysis
Kernel Methods
Face Recognition
Efficiency
Pattern Recognition
Discrimination
Research

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

Zhang, B., Ko, H., & Gao, Y. (2007). Learning kernel subspace classifier. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4642 LNCS, pp. 299-308). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4642 LNCS).

Learning kernel subspace classifier. / Zhang, Bailing; Ko, Hanseok; Gao, Yongsheng.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4642 LNCS 2007. p. 299-308 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4642 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhang, B, Ko, H & Gao, Y 2007, Learning kernel subspace classifier. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 4642 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4642 LNCS, pp. 299-308, 2007 International Conference on Advances in Biometrics, ICB 2007, Seoul, Korea, Republic of, 07/8/27.
Zhang B, Ko H, Gao Y. Learning kernel subspace classifier. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4642 LNCS. 2007. p. 299-308. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Zhang, Bailing ; Ko, Hanseok ; Gao, Yongsheng. / Learning kernel subspace classifier. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4642 LNCS 2007. pp. 299-308 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{32234fc9b1714a638ad5c5cdd4ac9872,
title = "Learning kernel subspace classifier",
abstract = "Subspace classifiers are well-known in pattern recognition, which represent pattern classes by linear subspaces spanned by the class specific basis vectors through simple mathematical operations like SVD. Recently, kernel based subspace methods have been proposed to extend the functionalities by directly applying the Kernel Principal Component Analysis (KPCA). The projection variance in kernel space as applied in these earlier proposed kernel subspace methods, however, is not a trustworthy criteria for class discrimination and they simply fail in many recognition problems as we encountered in biometrics research. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in input space through the kernel subspace projection. While the pre-image methods aiming at finding an approximate pre-image for each input by minimization of the reconstruction error in kernel space, we emphasize the problem of how to estimate a kernel subspace as a model for a specific class. Using the occluded face recognition as examples, our experimental results demonstrated the efficiency of the proposed method.",
author = "Bailing Zhang and Hanseok Ko and Yongsheng Gao",
year = "2007",
month = "12",
day = "1",
language = "English",
isbn = "9783540745488",
volume = "4642 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "299--308",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Learning kernel subspace classifier

AU - Zhang, Bailing

AU - Ko, Hanseok

AU - Gao, Yongsheng

PY - 2007/12/1

Y1 - 2007/12/1

N2 - Subspace classifiers are well-known in pattern recognition, which represent pattern classes by linear subspaces spanned by the class specific basis vectors through simple mathematical operations like SVD. Recently, kernel based subspace methods have been proposed to extend the functionalities by directly applying the Kernel Principal Component Analysis (KPCA). The projection variance in kernel space as applied in these earlier proposed kernel subspace methods, however, is not a trustworthy criteria for class discrimination and they simply fail in many recognition problems as we encountered in biometrics research. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in input space through the kernel subspace projection. While the pre-image methods aiming at finding an approximate pre-image for each input by minimization of the reconstruction error in kernel space, we emphasize the problem of how to estimate a kernel subspace as a model for a specific class. Using the occluded face recognition as examples, our experimental results demonstrated the efficiency of the proposed method.

AB - Subspace classifiers are well-known in pattern recognition, which represent pattern classes by linear subspaces spanned by the class specific basis vectors through simple mathematical operations like SVD. Recently, kernel based subspace methods have been proposed to extend the functionalities by directly applying the Kernel Principal Component Analysis (KPCA). The projection variance in kernel space as applied in these earlier proposed kernel subspace methods, however, is not a trustworthy criteria for class discrimination and they simply fail in many recognition problems as we encountered in biometrics research. We address this issue by proposing a learning kernel subspace classifier which attempts to reconstruct data in input space through the kernel subspace projection. While the pre-image methods aiming at finding an approximate pre-image for each input by minimization of the reconstruction error in kernel space, we emphasize the problem of how to estimate a kernel subspace as a model for a specific class. Using the occluded face recognition as examples, our experimental results demonstrated the efficiency of the proposed method.

UR - http://www.scopus.com/inward/record.url?scp=37849022817&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=37849022817&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:37849022817

SN - 9783540745488

VL - 4642 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 299

EP - 308

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -