Weighted discriminant analysis and kernel ridge regression metric learning for face verification

Siew Chin Chong; Andrew Beng Jin Teoh; Thian Song Ong

doi:10.1007/978-3-319-46672-9_45

Weighted discriminant analysis and kernel ridge regression metric learning for face verification

Siew Chin Chong, Andrew Beng Jin Teoh, Thian Song Ong

Department of Electrical and Electronic Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

A new formulation of metric learning is introduced by assimilating the kernel ridge regression (KRR) and weighted side-information linear discriminant analysis (WSILD) to enjoy the best of both worlds for unconstrained face verification task. To be specific, we formulate a doublet constrained metric learning problem by means of a second degree polynomial kernel function. The said metric learning problem can be solved analytically for Mahalanobis distance metric due to simplistic nature of KRR in which we named KRRML. In addition, the WSILD further enhances the learned Mahalanobis distance metric by leveraging the within-class and between-class scatter matrix of doublets. We evaluate the proposed method with Labeled Faces in the Wild database, a large benchmark dataset targeted for unconstrained face verification. The promising result attests the robustness and feasibility of the proposed method.

Original language	English
Title of host publication	Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings
Editors	Seiichi Ozawa, Kazushi Ikeda, Derong Liu, Akira Hirose, Kenji Doya, Minho Lee
Publisher	Springer Verlag
Pages	401-410
Number of pages	10
ISBN (Print)	9783319466712
DOIs	https://doi.org/10.1007/978-3-319-46672-9_45
Publication status	Published - 2016 Jan 1
Event	23rd International Conference on Neural Information Processing, ICONIP 2016 - Kyoto, Japan Duration: 2016 Oct 16 → 2016 Oct 21

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9948 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	23rd International Conference on Neural Information Processing, ICONIP 2016
Country	Japan
City	Kyoto
Period	16/10/16 → 16/10/21

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-319-46672-9_45

Fingerprint Dive into the research topics of 'Weighted discriminant analysis and kernel ridge regression metric learning for face verification'. Together they form a unique fingerprint.

Cite this

Chong, S. C., Teoh, A. B. J., & Ong, T. S. (2016). Weighted discriminant analysis and kernel ridge regression metric learning for face verification. In S. Ozawa, K. Ikeda, D. Liu, A. Hirose, K. Doya, & M. Lee (Eds.), Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings (pp. 401-410). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9948 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46672-9_45

Chong, Siew Chin ; Teoh, Andrew Beng Jin ; Ong, Thian Song. / Weighted discriminant analysis and kernel ridge regression metric learning for face verification. Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings. editor / Seiichi Ozawa ; Kazushi Ikeda ; Derong Liu ; Akira Hirose ; Kenji Doya ; Minho Lee. Springer Verlag, 2016. pp. 401-410 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{2974c246476d426b85d956ac61a9b84a,

title = "Weighted discriminant analysis and kernel ridge regression metric learning for face verification",

abstract = "A new formulation of metric learning is introduced by assimilating the kernel ridge regression (KRR) and weighted side-information linear discriminant analysis (WSILD) to enjoy the best of both worlds for unconstrained face verification task. To be specific, we formulate a doublet constrained metric learning problem by means of a second degree polynomial kernel function. The said metric learning problem can be solved analytically for Mahalanobis distance metric due to simplistic nature of KRR in which we named KRRML. In addition, the WSILD further enhances the learned Mahalanobis distance metric by leveraging the within-class and between-class scatter matrix of doublets. We evaluate the proposed method with Labeled Faces in the Wild database, a large benchmark dataset targeted for unconstrained face verification. The promising result attests the robustness and feasibility of the proposed method.",

author = "Chong, {Siew Chin} and Teoh, {Andrew Beng Jin} and Ong, {Thian Song}",

year = "2016",

month = jan,

day = "1",

doi = "10.1007/978-3-319-46672-9_45",

language = "English",

isbn = "9783319466712",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "401--410",

editor = "Seiichi Ozawa and Kazushi Ikeda and Derong Liu and Akira Hirose and Kenji Doya and Minho Lee",

booktitle = "Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings",

address = "Germany",

note = "23rd International Conference on Neural Information Processing, ICONIP 2016 ; Conference date: 16-10-2016 Through 21-10-2016",

}

Chong, SC, Teoh, ABJ & Ong, TS 2016, Weighted discriminant analysis and kernel ridge regression metric learning for face verification. in S Ozawa, K Ikeda, D Liu, A Hirose, K Doya & M Lee (eds), Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9948 LNCS, Springer Verlag, pp. 401-410, 23rd International Conference on Neural Information Processing, ICONIP 2016, Kyoto, Japan, 16/10/16. https://doi.org/10.1007/978-3-319-46672-9_45

Weighted discriminant analysis and kernel ridge regression metric learning for face verification. / Chong, Siew Chin; Teoh, Andrew Beng Jin; Ong, Thian Song.

Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings. ed. / Seiichi Ozawa; Kazushi Ikeda; Derong Liu; Akira Hirose; Kenji Doya; Minho Lee. Springer Verlag, 2016. p. 401-410 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9948 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Weighted discriminant analysis and kernel ridge regression metric learning for face verification

AU - Chong, Siew Chin

AU - Teoh, Andrew Beng Jin

AU - Ong, Thian Song

PY - 2016/1/1

Y1 - 2016/1/1

N2 - A new formulation of metric learning is introduced by assimilating the kernel ridge regression (KRR) and weighted side-information linear discriminant analysis (WSILD) to enjoy the best of both worlds for unconstrained face verification task. To be specific, we formulate a doublet constrained metric learning problem by means of a second degree polynomial kernel function. The said metric learning problem can be solved analytically for Mahalanobis distance metric due to simplistic nature of KRR in which we named KRRML. In addition, the WSILD further enhances the learned Mahalanobis distance metric by leveraging the within-class and between-class scatter matrix of doublets. We evaluate the proposed method with Labeled Faces in the Wild database, a large benchmark dataset targeted for unconstrained face verification. The promising result attests the robustness and feasibility of the proposed method.

AB - A new formulation of metric learning is introduced by assimilating the kernel ridge regression (KRR) and weighted side-information linear discriminant analysis (WSILD) to enjoy the best of both worlds for unconstrained face verification task. To be specific, we formulate a doublet constrained metric learning problem by means of a second degree polynomial kernel function. The said metric learning problem can be solved analytically for Mahalanobis distance metric due to simplistic nature of KRR in which we named KRRML. In addition, the WSILD further enhances the learned Mahalanobis distance metric by leveraging the within-class and between-class scatter matrix of doublets. We evaluate the proposed method with Labeled Faces in the Wild database, a large benchmark dataset targeted for unconstrained face verification. The promising result attests the robustness and feasibility of the proposed method.

UR - http://www.scopus.com/inward/record.url?scp=84992553675&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84992553675&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46672-9_45

DO - 10.1007/978-3-319-46672-9_45

M3 - Conference contribution

AN - SCOPUS:84992553675

SN - 9783319466712

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

SP - 401

EP - 410

BT - Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings

A2 - Ozawa, Seiichi

A2 - Ikeda, Kazushi

A2 - Liu, Derong

A2 - Hirose, Akira

A2 - Doya, Kenji

A2 - Lee, Minho

PB - Springer Verlag

T2 - 23rd International Conference on Neural Information Processing, ICONIP 2016

Y2 - 16 October 2016 through 21 October 2016

ER -

Chong SC, Teoh ABJ, Ong TS. Weighted discriminant analysis and kernel ridge regression metric learning for face verification. In Ozawa S, Ikeda K, Liu D, Hirose A, Doya K, Lee M, editors, Neural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings. Springer Verlag. 2016. p. 401-410. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46672-9_45