Reducing Variance of Committee Prediction with Resampling Techniques

Bambang Parmanto, Paul W. Munro, Howard Doyle

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

Algorithms for reducing variance in neural network prediction using committee and resampling techniques (bootstrap and cross-validation) are presented. Their effectiveness is tested on data sets with different levels of noise and on medical diagnosis data sets. The methods are most effective when the noise level in the data is high or the size of the learning set is small, which consequently produces high variance. The algorithms will not be of much help in cases where the error of prediction is mainly due to bias. An increase in bias is observed due to smaller effective learning size in the bootstrap and cross-validation committee. The impact of increased bias on the performance ranges from negligible to completely undermining the benefit of reducing the variance.

Original languageEnglish (US)
Pages (from-to)405-425
Number of pages21
JournalConnection Science
Volume8
Issue number3-4
DOIs
StatePublished - Dec 1996
Externally publishedYes

Fingerprint

Resampling
Cross-validation
Bootstrap
Prediction
Neural networks
Neural Networks
Range of data
Learning

Keywords

  • Bias-variance
  • Bootstrap
  • Committee
  • Cross-validation
  • Ensemble
  • Regularization
  • Resampling techniques

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Artificial Intelligence
  • Theoretical Computer Science
  • Computational Theory and Mathematics

Cite this

Reducing Variance of Committee Prediction with Resampling Techniques. / Parmanto, Bambang; Munro, Paul W.; Doyle, Howard.

In: Connection Science, Vol. 8, No. 3-4, 12.1996, p. 405-425.

Research output: Contribution to journalArticle

Parmanto, Bambang ; Munro, Paul W. ; Doyle, Howard. / Reducing Variance of Committee Prediction with Resampling Techniques. In: Connection Science. 1996 ; Vol. 8, No. 3-4. pp. 405-425.
@article{3d39f4c9075348a5a652dd64efdb4400,
title = "Reducing Variance of Committee Prediction with Resampling Techniques",
abstract = "Algorithms for reducing variance in neural network prediction using committee and resampling techniques (bootstrap and cross-validation) are presented. Their effectiveness is tested on data sets with different levels of noise and on medical diagnosis data sets. The methods are most effective when the noise level in the data is high or the size of the learning set is small, which consequently produces high variance. The algorithms will not be of much help in cases where the error of prediction is mainly due to bias. An increase in bias is observed due to smaller effective learning size in the bootstrap and cross-validation committee. The impact of increased bias on the performance ranges from negligible to completely undermining the benefit of reducing the variance.",
keywords = "Bias-variance, Bootstrap, Committee, Cross-validation, Ensemble, Regularization, Resampling techniques",
author = "Bambang Parmanto and Munro, {Paul W.} and Howard Doyle",
year = "1996",
month = "12",
doi = "10.1080/095400996116848",
language = "English (US)",
volume = "8",
pages = "405--425",
journal = "Connection Science",
issn = "0954-0091",
publisher = "Taylor and Francis AS",
number = "3-4",

}

TY - JOUR

T1 - Reducing Variance of Committee Prediction with Resampling Techniques

AU - Parmanto, Bambang

AU - Munro, Paul W.

AU - Doyle, Howard

PY - 1996/12

Y1 - 1996/12

N2 - Algorithms for reducing variance in neural network prediction using committee and resampling techniques (bootstrap and cross-validation) are presented. Their effectiveness is tested on data sets with different levels of noise and on medical diagnosis data sets. The methods are most effective when the noise level in the data is high or the size of the learning set is small, which consequently produces high variance. The algorithms will not be of much help in cases where the error of prediction is mainly due to bias. An increase in bias is observed due to smaller effective learning size in the bootstrap and cross-validation committee. The impact of increased bias on the performance ranges from negligible to completely undermining the benefit of reducing the variance.

AB - Algorithms for reducing variance in neural network prediction using committee and resampling techniques (bootstrap and cross-validation) are presented. Their effectiveness is tested on data sets with different levels of noise and on medical diagnosis data sets. The methods are most effective when the noise level in the data is high or the size of the learning set is small, which consequently produces high variance. The algorithms will not be of much help in cases where the error of prediction is mainly due to bias. An increase in bias is observed due to smaller effective learning size in the bootstrap and cross-validation committee. The impact of increased bias on the performance ranges from negligible to completely undermining the benefit of reducing the variance.

KW - Bias-variance

KW - Bootstrap

KW - Committee

KW - Cross-validation

KW - Ensemble

KW - Regularization

KW - Resampling techniques

UR - http://www.scopus.com/inward/record.url?scp=0030352275&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030352275&partnerID=8YFLogxK

U2 - 10.1080/095400996116848

DO - 10.1080/095400996116848

M3 - Article

VL - 8

SP - 405

EP - 425

JO - Connection Science

JF - Connection Science

SN - 0954-0091

IS - 3-4

ER -