A comparative study on mirror image learning and ALSM

Tetsushi Wakabayashi; Meng Shi; Wataru Oyama; Fumitaka Kimura

doi:10.1109/IWFHR.2002.1030901

A comparative study on mirror image learning and ALSM

Tetsushi Wakabayashi, Meng Shi, Wataru Oyama, Fumitaka Kimura

Real world robotics

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

1 Citation (Scopus)

Abstract

In this paper, the effectiveness of a corrective learning algorithm MIL (mirror image learning) is comparatively studied with that of ALSM (average learning subspace method). Both MIL and ALSM were proposed to improve the learning effectiveness of class conditional distributions. While the ALSM modifies the basis vectors of a subspace by subtracting the autocorrelation matrix for counter classes from the one of its own class, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes to increases the size of the learning sample of the other class. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CDROMI. Experimental results show that the recognition rate of the subspace method is improved from 99.05% to 99.37% by ALSM and to 99.39% by MIL, respectively. Furthermore, the recognition rate of the projection distance method is improved from 99.13% to 99.35% by ALSM and to 99.44% by MIL.

Original language	English
Title of host publication	Proceedings - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002
Pages	151-156
Number of pages	6
DOIs	https://doi.org/10.1109/IWFHR.2002.1030901
Publication status	Published - Dec 1 2002
Event	8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002 - Ontario, ON, Canada Duration: Aug 6 2002 → Aug 8 2002

Publication series

Name	Proceedings - International Workshop on Frontiers in Handwriting Recognition, IWFHR
ISSN (Print)	1550-5235

Other

Other	8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002
Country	Canada
City	Ontario, ON
Period	8/6/02 → 8/8/02

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition

Access to Document

10.1109/IWFHR.2002.1030901

Fingerprint Dive into the research topics of 'A comparative study on mirror image learning and ALSM'. Together they form a unique fingerprint.

Cite this

A comparative study on mirror image learning and ALSM. / Wakabayashi, Tetsushi; Shi, Meng; Oyama, Wataru; Kimura, Fumitaka.

Proceedings - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002. 2002. p. 151-156 1030901 (Proceedings - International Workshop on Frontiers in Handwriting Recognition, IWFHR).

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

Wakabayashi, T, Shi, M, Oyama, W & Kimura, F 2002, A comparative study on mirror image learning and ALSM. in Proceedings - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002., 1030901, Proceedings - International Workshop on Frontiers in Handwriting Recognition, IWFHR, pp. 151-156, 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002, Ontario, ON, Canada, 8/6/02. https://doi.org/10.1109/IWFHR.2002.1030901

@inproceedings{fa2108b486a245c0b0b4eeb0eb52d500,

title = "A comparative study on mirror image learning and ALSM",

abstract = "In this paper, the effectiveness of a corrective learning algorithm MIL (mirror image learning) is comparatively studied with that of ALSM (average learning subspace method). Both MIL and ALSM were proposed to improve the learning effectiveness of class conditional distributions. While the ALSM modifies the basis vectors of a subspace by subtracting the autocorrelation matrix for counter classes from the one of its own class, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes to increases the size of the learning sample of the other class. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CDROMI. Experimental results show that the recognition rate of the subspace method is improved from 99.05% to 99.37% by ALSM and to 99.39% by MIL, respectively. Furthermore, the recognition rate of the projection distance method is improved from 99.13% to 99.35% by ALSM and to 99.44% by MIL.",

author = "Tetsushi Wakabayashi and Meng Shi and Wataru Oyama and Fumitaka Kimura",

year = "2002",

month = dec,

day = "1",

doi = "10.1109/IWFHR.2002.1030901",

language = "English",

isbn = "0769516920",

series = "Proceedings - International Workshop on Frontiers in Handwriting Recognition, IWFHR",

pages = "151--156",

booktitle = "Proceedings - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002",

note = "8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002 ; Conference date: 06-08-2002 Through 08-08-2002",

}

TY - GEN

T1 - A comparative study on mirror image learning and ALSM

AU - Wakabayashi, Tetsushi

AU - Shi, Meng

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) is comparatively studied with that of ALSM (average learning subspace method). Both MIL and ALSM were proposed to improve the learning effectiveness of class conditional distributions. While the ALSM modifies the basis vectors of a subspace by subtracting the autocorrelation matrix for counter classes from the one of its own class, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes to increases the size of the learning sample of the other class. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CDROMI. Experimental results show that the recognition rate of the subspace method is improved from 99.05% to 99.37% by ALSM and to 99.39% by MIL, respectively. Furthermore, the recognition rate of the projection distance method is improved from 99.13% to 99.35% by ALSM and to 99.44% by MIL.

AB - In this paper, the effectiveness of a corrective learning algorithm MIL (mirror image learning) is comparatively studied with that of ALSM (average learning subspace method). Both MIL and ALSM were proposed to improve the learning effectiveness of class conditional distributions. While the ALSM modifies the basis vectors of a subspace by subtracting the autocorrelation matrix for counter classes from the one of its own class, the MIL generates a mirror image of a pattern which belongs to one of a pair of confusing classes to increases the size of the learning sample of the other class. The performance of two algorithms is evaluated on handwritten numeral recognition test for IPTP CDROMI. Experimental results show that the recognition rate of the subspace method is improved from 99.05% to 99.37% by ALSM and to 99.39% by MIL, respectively. Furthermore, the recognition rate of the projection distance method is improved from 99.13% to 99.35% by ALSM and to 99.44% by MIL.

UR - http://www.scopus.com/inward/record.url?scp=71249161793&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=71249161793&partnerID=8YFLogxK

U2 - 10.1109/IWFHR.2002.1030901

DO - 10.1109/IWFHR.2002.1030901

M3 - Conference contribution

AN - SCOPUS:71249161793

SN - 0769516920

SN - 9780769516929

T3 - Proceedings - International Workshop on Frontiers in Handwriting Recognition, IWFHR

SP - 151

EP - 156

BT - Proceedings - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002

T2 - 8th International Workshop on Frontiers in Handwriting Recognition, IWFHR 2002

Y2 - 6 August 2002 through 8 August 2002

ER -