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

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

Research output: Chapter in Book/Report/Conference proceedingConference 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 languageEnglish
Title of host publicationNeural Information Processing - 23rd International Conference, ICONIP 2016, Proceedings
EditorsSeiichi Ozawa, Kazushi Ikeda, Derong Liu, Akira Hirose, Kenji Doya, Minho Lee
PublisherSpringer Verlag
Pages401-410
Number of pages10
ISBN (Print)9783319466712
DOIs
Publication statusPublished - 2016 Jan 1
Event23rd International Conference on Neural Information Processing, ICONIP 2016 - Kyoto, Japan
Duration: 2016 Oct 162016 Oct 21

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9948 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other23rd International Conference on Neural Information Processing, ICONIP 2016
CountryJapan
CityKyoto
Period16/10/1616/10/21

Fingerprint

Kernel Regression
Ridge Regression
Discriminant analysis
Discriminant Analysis
Mahalanobis Distance
Side Information
Distance Metric
Face
Metric
Polynomials
Scatter
Polynomial function
Kernel Function
Benchmark
Robustness
Formulation
Evaluate
Learning
Class

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

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 = "1",
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",

}

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 proceedingConference 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

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