Fast statistical learning algorithm for feature generation

Minoru Fukumi, Stephen Karungaru, Satoru Tsuge, Miyoko Nakano, Takuya Akashi, Yasue Mitsukura

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

2 Citations (Scopus)

Abstract

This paper presents an improved statistical learning algorithm for feature generation in pattern recognition and signal processing. It is approximately derived from geometrical interpretation of the Fisher linear discriminant analysis (FLDA). The principal component analysis (PCA) is popular for data compression and feature extraction. Furthermore, iterative learning algorithms for obtaining eigenvectors in PCA have been presented in such fields. Their effectiveness has been demonstrated in many applications. However, recently FLDA has been often used in many fields, especially face image recognition. The drawback of FLDA is a long computational time based on a large-sized covariance matrix and the issue that the within-class covariance matrix is usually singular. Generally, in FLDA, the inverse matrix of the within-class covariance matrix cannot be obtained, since data dimension is generally higher than the number of data and then it includes many zero eigenvalues. In order to overcome this difficulty, a new iterative feature generation method, the simple-FLDA was proposed by authors. In this paper, further improvement is introduced into the simple-FLDA and its effectiveness is demonstrated for preliminary personal identification problem.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages91-97
Number of pages7
Volume4694 LNAI
EditionPART 3
Publication statusPublished - 2007
Externally publishedYes
Event11th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2007, and 17th Italian Workshop on Neural Networks, WIRN 2007 - Vietri sul Mare, Italy
Duration: 2007 Sep 122007 Sep 14

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 3
Volume4694 LNAI
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other11th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2007, and 17th Italian Workshop on Neural Networks, WIRN 2007
CountryItaly
CityVietri sul Mare
Period07/9/1207/9/14

Fingerprint

Statistical Learning
Discriminant Analysis
Discriminant analysis
Learning algorithms
Learning Algorithm
Learning
Covariance matrix
Principal Component Analysis
Principal component analysis
Data Compression
Image recognition
Image Recognition
Inverse matrix
Data compression
Identification Problem
Eigenvalues and eigenfunctions
Iterative Algorithm
Eigenvector
Feature Extraction
Pattern Recognition

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

Fukumi, M., Karungaru, S., Tsuge, S., Nakano, M., Akashi, T., & Mitsukura, Y. (2007). Fast statistical learning algorithm for feature generation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 3 ed., Vol. 4694 LNAI, pp. 91-97). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4694 LNAI, No. PART 3).

Fast statistical learning algorithm for feature generation. / Fukumi, Minoru; Karungaru, Stephen; Tsuge, Satoru; Nakano, Miyoko; Akashi, Takuya; Mitsukura, Yasue.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4694 LNAI PART 3. ed. 2007. p. 91-97 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4694 LNAI, No. PART 3).

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

Fukumi, M, Karungaru, S, Tsuge, S, Nakano, M, Akashi, T & Mitsukura, Y 2007, Fast statistical learning algorithm for feature generation. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 3 edn, vol. 4694 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 3, vol. 4694 LNAI, pp. 91-97, 11th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2007, and 17th Italian Workshop on Neural Networks, WIRN 2007, Vietri sul Mare, Italy, 07/9/12.
Fukumi M, Karungaru S, Tsuge S, Nakano M, Akashi T, Mitsukura Y. Fast statistical learning algorithm for feature generation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 3 ed. Vol. 4694 LNAI. 2007. p. 91-97. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3).
Fukumi, Minoru ; Karungaru, Stephen ; Tsuge, Satoru ; Nakano, Miyoko ; Akashi, Takuya ; Mitsukura, Yasue. / Fast statistical learning algorithm for feature generation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4694 LNAI PART 3. ed. 2007. pp. 91-97 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3).
@inproceedings{a0a8d8db86eb4948a3b23081c439e096,
title = "Fast statistical learning algorithm for feature generation",
abstract = "This paper presents an improved statistical learning algorithm for feature generation in pattern recognition and signal processing. It is approximately derived from geometrical interpretation of the Fisher linear discriminant analysis (FLDA). The principal component analysis (PCA) is popular for data compression and feature extraction. Furthermore, iterative learning algorithms for obtaining eigenvectors in PCA have been presented in such fields. Their effectiveness has been demonstrated in many applications. However, recently FLDA has been often used in many fields, especially face image recognition. The drawback of FLDA is a long computational time based on a large-sized covariance matrix and the issue that the within-class covariance matrix is usually singular. Generally, in FLDA, the inverse matrix of the within-class covariance matrix cannot be obtained, since data dimension is generally higher than the number of data and then it includes many zero eigenvalues. In order to overcome this difficulty, a new iterative feature generation method, the simple-FLDA was proposed by authors. In this paper, further improvement is introduced into the simple-FLDA and its effectiveness is demonstrated for preliminary personal identification problem.",
author = "Minoru Fukumi and Stephen Karungaru and Satoru Tsuge and Miyoko Nakano and Takuya Akashi and Yasue Mitsukura",
year = "2007",
language = "English",
isbn = "9783540748281",
volume = "4694 LNAI",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 3",
pages = "91--97",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 3",

}

TY - GEN

T1 - Fast statistical learning algorithm for feature generation

AU - Fukumi, Minoru

AU - Karungaru, Stephen

AU - Tsuge, Satoru

AU - Nakano, Miyoko

AU - Akashi, Takuya

AU - Mitsukura, Yasue

PY - 2007

Y1 - 2007

N2 - This paper presents an improved statistical learning algorithm for feature generation in pattern recognition and signal processing. It is approximately derived from geometrical interpretation of the Fisher linear discriminant analysis (FLDA). The principal component analysis (PCA) is popular for data compression and feature extraction. Furthermore, iterative learning algorithms for obtaining eigenvectors in PCA have been presented in such fields. Their effectiveness has been demonstrated in many applications. However, recently FLDA has been often used in many fields, especially face image recognition. The drawback of FLDA is a long computational time based on a large-sized covariance matrix and the issue that the within-class covariance matrix is usually singular. Generally, in FLDA, the inverse matrix of the within-class covariance matrix cannot be obtained, since data dimension is generally higher than the number of data and then it includes many zero eigenvalues. In order to overcome this difficulty, a new iterative feature generation method, the simple-FLDA was proposed by authors. In this paper, further improvement is introduced into the simple-FLDA and its effectiveness is demonstrated for preliminary personal identification problem.

AB - This paper presents an improved statistical learning algorithm for feature generation in pattern recognition and signal processing. It is approximately derived from geometrical interpretation of the Fisher linear discriminant analysis (FLDA). The principal component analysis (PCA) is popular for data compression and feature extraction. Furthermore, iterative learning algorithms for obtaining eigenvectors in PCA have been presented in such fields. Their effectiveness has been demonstrated in many applications. However, recently FLDA has been often used in many fields, especially face image recognition. The drawback of FLDA is a long computational time based on a large-sized covariance matrix and the issue that the within-class covariance matrix is usually singular. Generally, in FLDA, the inverse matrix of the within-class covariance matrix cannot be obtained, since data dimension is generally higher than the number of data and then it includes many zero eigenvalues. In order to overcome this difficulty, a new iterative feature generation method, the simple-FLDA was proposed by authors. In this paper, further improvement is introduced into the simple-FLDA and its effectiveness is demonstrated for preliminary personal identification problem.

UR - http://www.scopus.com/inward/record.url?scp=38049107774&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=38049107774&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:38049107774

SN - 9783540748281

VL - 4694 LNAI

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

SP - 91

EP - 97

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

ER -