Comparative study on Mirror Image Learning (MIL) and GLVQ

Meng Shi, Tetsushi Wakabayashi, Wataru Oyama, Fumitaka Kimura

Research output: Contribution to journalArticle

Abstract

In this paper the effectiveness of a corrective learning algorithm MIL (Mirror Image Learning) [1], [2] is comparatively studied with that of GLVQ (Generalized Learning Vector Quantization) [3]. Both MIL and GLVQ were proposed to improve the learning effectiveness beyond the limitation due to independent estimation of class conditional distributions. While the GLVQ modifies the representative vectors of a pair of confusing classes when recognizing each learning pattern, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes and increases the size of the learning sample to update the distribution parameters. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CD-ROMI [4]. Experimental results show that the recognition rate of projection distance classifier is improved from 99.31% to 99.40% by GLVQ and to 99.50% by MIL, respectively.

Original languageEnglish
Pages (from-to)248-252
Number of pages5
JournalProceedings - International Conference on Pattern Recognition
Volume16
Issue number2
Publication statusPublished - Dec 1 2002

Fingerprint

Vector quantization
Mirrors
Learning algorithms
Classifiers

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Comparative study on Mirror Image Learning (MIL) and GLVQ. / Shi, Meng; Wakabayashi, Tetsushi; Oyama, Wataru; Kimura, Fumitaka.

In: Proceedings - International Conference on Pattern Recognition, Vol. 16, No. 2, 01.12.2002, p. 248-252.

Research output: Contribution to journalArticle

Shi, M, Wakabayashi, T, Oyama, W & Kimura, F 2002, 'Comparative study on Mirror Image Learning (MIL) and GLVQ', Proceedings - International Conference on Pattern Recognition, vol. 16, no. 2, pp. 248-252.
Shi, Meng ; Wakabayashi, Tetsushi ; Oyama, Wataru ; Kimura, Fumitaka. / Comparative study on Mirror Image Learning (MIL) and GLVQ. In: Proceedings - International Conference on Pattern Recognition. 2002 ; Vol. 16, No. 2. pp. 248-252.
@article{07889047324d43738931ff56c9d82261,
title = "Comparative study on Mirror Image Learning (MIL) and GLVQ",
abstract = "In this paper the effectiveness of a corrective learning algorithm MIL (Mirror Image Learning) [1], [2] is comparatively studied with that of GLVQ (Generalized Learning Vector Quantization) [3]. Both MIL and GLVQ were proposed to improve the learning effectiveness beyond the limitation due to independent estimation of class conditional distributions. While the GLVQ modifies the representative vectors of a pair of confusing classes when recognizing each learning pattern, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes and increases the size of the learning sample to update the distribution parameters. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CD-ROMI [4]. Experimental results show that the recognition rate of projection distance classifier is improved from 99.31{\%} to 99.40{\%} by GLVQ and to 99.50{\%} by MIL, respectively.",
author = "Meng Shi and Tetsushi Wakabayashi and Wataru Oyama and Fumitaka Kimura",
year = "2002",
month = "12",
day = "1",
language = "English",
volume = "16",
pages = "248--252",
journal = "Proceedings - International Conference on Pattern Recognition",
issn = "1051-4651",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "2",

}

TY - JOUR

T1 - Comparative study on Mirror Image Learning (MIL) and GLVQ

AU - Shi, Meng

AU - Wakabayashi, Tetsushi

AU - Oyama, Wataru

AU - Kimura, Fumitaka

PY - 2002/12/1

Y1 - 2002/12/1

N2 - In this paper the effectiveness of a corrective learning algorithm MIL (Mirror Image Learning) [1], [2] is comparatively studied with that of GLVQ (Generalized Learning Vector Quantization) [3]. Both MIL and GLVQ were proposed to improve the learning effectiveness beyond the limitation due to independent estimation of class conditional distributions. While the GLVQ modifies the representative vectors of a pair of confusing classes when recognizing each learning pattern, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes and increases the size of the learning sample to update the distribution parameters. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CD-ROMI [4]. Experimental results show that the recognition rate of projection distance classifier is improved from 99.31% to 99.40% by GLVQ and to 99.50% by MIL, respectively.

AB - In this paper the effectiveness of a corrective learning algorithm MIL (Mirror Image Learning) [1], [2] is comparatively studied with that of GLVQ (Generalized Learning Vector Quantization) [3]. Both MIL and GLVQ were proposed to improve the learning effectiveness beyond the limitation due to independent estimation of class conditional distributions. While the GLVQ modifies the representative vectors of a pair of confusing classes when recognizing each learning pattern, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes and increases the size of the learning sample to update the distribution parameters. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CD-ROMI [4]. Experimental results show that the recognition rate of projection distance classifier is improved from 99.31% to 99.40% by GLVQ and to 99.50% by MIL, respectively.

UR - http://www.scopus.com/inward/record.url?scp=33751578387&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33751578387&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:33751578387

VL - 16

SP - 248

EP - 252

JO - Proceedings - International Conference on Pattern Recognition

JF - Proceedings - International Conference on Pattern Recognition

SN - 1051-4651

IS - 2

ER -