Combination of multiple classifiers by minimizing the upper bound of Bayes error rate for unconstrained handwritten numeral recognition

Hee Joong Kang, Seong Whan Lee

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

In order to raise a class discrimination power by the combination of multiple classifiers, the upper bound of Bayes error rate which is bounded by the conditional entropy of a class and decisions should be minimized. Based on the minimization of the upper bound of the Bayes error rate, Wang and Wong proposed only a tree dependence approximation scheme of a high-dimensional probability distribution composed of a class and patterns. This paper extends such a tree dependence approximation scheme to higher order dependency for improving the classification performance and thus optimally approximates the high-dimensional probability distribution with a product of low-dimensional distributions. And then, a new combination method by the proposed approximation scheme is presented and' evaluated with classifiers recognizing unconstrained handwritten numerals.

Original languageEnglish
Pages (from-to)395-413
Number of pages19
JournalInternational Journal of Pattern Recognition and Artificial Intelligence
Volume19
Issue number3
DOIs
Publication statusPublished - 2005 May 1

Fingerprint

Probability distributions
Classifiers
Entropy

Keywords

  • Approximation scheme: mutual information
  • Bayes error rate
  • Combination of multiple classifiers
  • Dependency
  • Handwritten numeral recognition

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Artificial Intelligence
  • Computer Vision and Pattern Recognition

Cite this

@article{fe4dca262b8b42069be1747ba07d6f62,
title = "Combination of multiple classifiers by minimizing the upper bound of Bayes error rate for unconstrained handwritten numeral recognition",
abstract = "In order to raise a class discrimination power by the combination of multiple classifiers, the upper bound of Bayes error rate which is bounded by the conditional entropy of a class and decisions should be minimized. Based on the minimization of the upper bound of the Bayes error rate, Wang and Wong proposed only a tree dependence approximation scheme of a high-dimensional probability distribution composed of a class and patterns. This paper extends such a tree dependence approximation scheme to higher order dependency for improving the classification performance and thus optimally approximates the high-dimensional probability distribution with a product of low-dimensional distributions. And then, a new combination method by the proposed approximation scheme is presented and' evaluated with classifiers recognizing unconstrained handwritten numerals.",
keywords = "Approximation scheme: mutual information, Bayes error rate, Combination of multiple classifiers, Dependency, Handwritten numeral recognition",
author = "Kang, {Hee Joong} and Lee, {Seong Whan}",
year = "2005",
month = "5",
day = "1",
doi = "10.1142/S0218001405004101",
language = "English",
volume = "19",
pages = "395--413",
journal = "International Journal of Pattern Recognition and Artificial Intelligence",
issn = "0218-0014",
publisher = "World Scientific Publishing Co. Pte Ltd",
number = "3",

}

TY - JOUR

T1 - Combination of multiple classifiers by minimizing the upper bound of Bayes error rate for unconstrained handwritten numeral recognition

AU - Kang, Hee Joong

AU - Lee, Seong Whan

PY - 2005/5/1

Y1 - 2005/5/1

N2 - In order to raise a class discrimination power by the combination of multiple classifiers, the upper bound of Bayes error rate which is bounded by the conditional entropy of a class and decisions should be minimized. Based on the minimization of the upper bound of the Bayes error rate, Wang and Wong proposed only a tree dependence approximation scheme of a high-dimensional probability distribution composed of a class and patterns. This paper extends such a tree dependence approximation scheme to higher order dependency for improving the classification performance and thus optimally approximates the high-dimensional probability distribution with a product of low-dimensional distributions. And then, a new combination method by the proposed approximation scheme is presented and' evaluated with classifiers recognizing unconstrained handwritten numerals.

AB - In order to raise a class discrimination power by the combination of multiple classifiers, the upper bound of Bayes error rate which is bounded by the conditional entropy of a class and decisions should be minimized. Based on the minimization of the upper bound of the Bayes error rate, Wang and Wong proposed only a tree dependence approximation scheme of a high-dimensional probability distribution composed of a class and patterns. This paper extends such a tree dependence approximation scheme to higher order dependency for improving the classification performance and thus optimally approximates the high-dimensional probability distribution with a product of low-dimensional distributions. And then, a new combination method by the proposed approximation scheme is presented and' evaluated with classifiers recognizing unconstrained handwritten numerals.

KW - Approximation scheme: mutual information

KW - Bayes error rate

KW - Combination of multiple classifiers

KW - Dependency

KW - Handwritten numeral recognition

UR - http://www.scopus.com/inward/record.url?scp=18644367804&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=18644367804&partnerID=8YFLogxK

U2 - 10.1142/S0218001405004101

DO - 10.1142/S0218001405004101

M3 - Article

VL - 19

SP - 395

EP - 413

JO - International Journal of Pattern Recognition and Artificial Intelligence

JF - International Journal of Pattern Recognition and Artificial Intelligence

SN - 0218-0014

IS - 3

ER -