A discriminative and heteroscedastic linear feature transformation for multiclass classification

Hung Shin Lee, Hsin Min Wang, Berlin Chen

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

1 Citation (Scopus)

Abstract

This paper presents a novel discriminative feature transformation, named full-rank generalized likelihood ratio discriminant analysis (fGLRDA), on the grounds of the likelihood ratio test (LRT). fGLRDA attempts to seek a feature space, which is linearly isomorphic to the original n-dimensional feature space and is characterized by a full-rank ) (nxn) transformation matrix, under the assumption that all the class-discrimination information resides in a d-dimensional subspace ) (d < n), through making the most confusing situation, described by the null hypothesis, as unlikely as possible to happen without the homoscedastic assumption on class distributions. Our experimental results demonstrate that fGLRDA can yield moderate performance improvements over other existing methods, such as linear discriminant analysis (LDA) for the speaker identification task.

Original languageEnglish
Title of host publicationProceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010
Pages690-693
Number of pages4
DOIs
Publication statusPublished - 2010 Nov 18
Event2010 20th International Conference on Pattern Recognition, ICPR 2010 - Istanbul, Turkey
Duration: 2010 Aug 232010 Aug 26

Publication series

NameProceedings - International Conference on Pattern Recognition
ISSN (Print)1051-4651

Other

Other2010 20th International Conference on Pattern Recognition, ICPR 2010
CountryTurkey
CityIstanbul
Period10/8/2310/8/26

Fingerprint

Discriminant analysis

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Lee, H. S., Wang, H. M., & Chen, B. (2010). A discriminative and heteroscedastic linear feature transformation for multiclass classification. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010 (pp. 690-693). [5595812] (Proceedings - International Conference on Pattern Recognition). https://doi.org/10.1109/ICPR.2010.174

A discriminative and heteroscedastic linear feature transformation for multiclass classification. / Lee, Hung Shin; Wang, Hsin Min; Chen, Berlin.

Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 690-693 5595812 (Proceedings - International Conference on Pattern Recognition).

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

Lee, HS, Wang, HM & Chen, B 2010, A discriminative and heteroscedastic linear feature transformation for multiclass classification. in Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010., 5595812, Proceedings - International Conference on Pattern Recognition, pp. 690-693, 2010 20th International Conference on Pattern Recognition, ICPR 2010, Istanbul, Turkey, 10/8/23. https://doi.org/10.1109/ICPR.2010.174
Lee HS, Wang HM, Chen B. A discriminative and heteroscedastic linear feature transformation for multiclass classification. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 690-693. 5595812. (Proceedings - International Conference on Pattern Recognition). https://doi.org/10.1109/ICPR.2010.174
Lee, Hung Shin ; Wang, Hsin Min ; Chen, Berlin. / A discriminative and heteroscedastic linear feature transformation for multiclass classification. Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. pp. 690-693 (Proceedings - International Conference on Pattern Recognition).
@inproceedings{659a07f2f709460dac81c68fc1695903,
title = "A discriminative and heteroscedastic linear feature transformation for multiclass classification",
abstract = "This paper presents a novel discriminative feature transformation, named full-rank generalized likelihood ratio discriminant analysis (fGLRDA), on the grounds of the likelihood ratio test (LRT). fGLRDA attempts to seek a feature space, which is linearly isomorphic to the original n-dimensional feature space and is characterized by a full-rank ) (nxn) transformation matrix, under the assumption that all the class-discrimination information resides in a d-dimensional subspace ) (d < n), through making the most confusing situation, described by the null hypothesis, as unlikely as possible to happen without the homoscedastic assumption on class distributions. Our experimental results demonstrate that fGLRDA can yield moderate performance improvements over other existing methods, such as linear discriminant analysis (LDA) for the speaker identification task.",
author = "Lee, {Hung Shin} and Wang, {Hsin Min} and Berlin Chen",
year = "2010",
month = "11",
day = "18",
doi = "10.1109/ICPR.2010.174",
language = "English",
isbn = "9780769541099",
series = "Proceedings - International Conference on Pattern Recognition",
pages = "690--693",
booktitle = "Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010",

}

TY - GEN

T1 - A discriminative and heteroscedastic linear feature transformation for multiclass classification

AU - Lee, Hung Shin

AU - Wang, Hsin Min

AU - Chen, Berlin

PY - 2010/11/18

Y1 - 2010/11/18

N2 - This paper presents a novel discriminative feature transformation, named full-rank generalized likelihood ratio discriminant analysis (fGLRDA), on the grounds of the likelihood ratio test (LRT). fGLRDA attempts to seek a feature space, which is linearly isomorphic to the original n-dimensional feature space and is characterized by a full-rank ) (nxn) transformation matrix, under the assumption that all the class-discrimination information resides in a d-dimensional subspace ) (d < n), through making the most confusing situation, described by the null hypothesis, as unlikely as possible to happen without the homoscedastic assumption on class distributions. Our experimental results demonstrate that fGLRDA can yield moderate performance improvements over other existing methods, such as linear discriminant analysis (LDA) for the speaker identification task.

AB - This paper presents a novel discriminative feature transformation, named full-rank generalized likelihood ratio discriminant analysis (fGLRDA), on the grounds of the likelihood ratio test (LRT). fGLRDA attempts to seek a feature space, which is linearly isomorphic to the original n-dimensional feature space and is characterized by a full-rank ) (nxn) transformation matrix, under the assumption that all the class-discrimination information resides in a d-dimensional subspace ) (d < n), through making the most confusing situation, described by the null hypothesis, as unlikely as possible to happen without the homoscedastic assumption on class distributions. Our experimental results demonstrate that fGLRDA can yield moderate performance improvements over other existing methods, such as linear discriminant analysis (LDA) for the speaker identification task.

UR - http://www.scopus.com/inward/record.url?scp=78149483651&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149483651&partnerID=8YFLogxK

U2 - 10.1109/ICPR.2010.174

DO - 10.1109/ICPR.2010.174

M3 - Conference contribution

AN - SCOPUS:78149483651

SN - 9780769541099

T3 - Proceedings - International Conference on Pattern Recognition

SP - 690

EP - 693

BT - Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010

ER -