Display Publications by [Year] [Type] [Topic]
Determining the performance bounds possible with a particular classifier or data set is often of great importance in pattern recognition applications. A previously introduced method for Bayes error estimation based on combining multiple classifiers outperforms more traditional estimates of this error in many instances. The accuracy of this estimate, however, relies on the correlation among the classifiers, a quantity that may be difficult to quantify precisely. Addressing this issue, we explore improvements to the ensemble based estimates of the Bayes error by considering information theoretic issues. More precisely, we use mutual information to determine a "similarity" measure between trained classifiers. This approach provides a more reliable similarity measure than error correlation, and leads to more accurate bounds on classification error. Application to an artificial problem with known Bayes error rate and a real world problem involving underwater acoustic data demonstrates the accuracy of this method.
(unavailable)
@inproceedings{tumer-bollacker_annie98,
author="K. Tumer and K. D. Bollacker and J. Ghosh",
title="A Mutual Information Based Ensemble Method
to Estimate the {B}ayes Error",
booktitle="Intelligent Engineering Systems through Artificial
Neural Networks",
editor ={C. Dagli et al.},
pages={17-22},
volume = {8},
publisher = {ASME Press},
abstract={Determining the performance bounds possible with a particular classifier or data set is often of great importance in pattern recognition applications. A previously introduced method for Bayes error estimation based on combining multiple classifiers outperforms more traditional estimates of this error in many instances. The accuracy of this estimate, however, relies on the correlation among the classifiers, a quantity that may be difficult to quantify precisely.
Addressing this issue, we explore improvements to the ensemble based estimates of the Bayes error by considering information theoretic issues. More precisely, we use mutual information to determine a "similarity" measure between trained classifiers. This approach provides a more reliable similarity measure than error correlation, and leads to more accurate bounds on classification error. Application to an artificial problem with known Bayes error rate and a real world problem involving underwater acoustic data demonstrates the accuracy of this method.},
bib2html_pubtype = {Refereed Conference Papers},
bib2html_rescat = {Bayes Error Estimation, Classifier Ensembles},
year={1998}
}
Generated by bib2html.pl (written by Patrick Riley ) on Wed Apr 01, 2020 17:39:43