Kagan Tumer's Publications

Display Publications by [Year] [Type] [Topic]


Limits to Performance Gains in Combined Neural Classifiers. K. Tumer and J. Ghosh. In Intelligent Engineering Systems through Artificial Neural Networks, pp. 419–424, St. Louis, November 1995.

Abstract

The performance of a single classifier is often inadequate in difficult classification problems. In such cases, several researchers have combined the outputs of multiple classifiers to obtain better performance. However, the amount of improvement possible through such combination techniques is generally not known. This article presents two approaches to estimating performance limits in hybrid networks. First, we present a framework that estimates Bayes error rates when linear combiners are used. Then we discuss a more general method that provides decision confidences and error bounds based on error types arising from the training data. The methods are illustrated for a difficult four class problem involving underwater acoustic data. For this data, we compute the single classifier and combiner classification performances, as well as the Bayes error rate and an error bound.

Download

(unavailable)

BibTeX Entry

@inproceedings{tumer-ghosh_annie95,
        author={K. Tumer and J. Ghosh},
        title="Limits to Performance Gains in Combined Neural Classifiers",
        booktitle="Intelligent Engineering Systems through 
		Artificial Neural Networks",
        volume ={7},
        pages ={419-424},
        address="St. Louis",
	month ={November},
	abstract={The performance of a single classifier is often inadequate in difficult classification problems. In such cases, several researchers have combined the outputs of multiple classifiers to obtain better performance. However, the amount of improvement possible through such combination techniques is generally not known. This article presents two approaches to estimating performance limits in hybrid networks. First, we present a framework that estimates Bayes error rates when linear combiners are used. Then we discuss a more general method that provides decision confidences and error bounds based on error types arising from the training data. The methods are illustrated for a difficult four class problem involving underwater acoustic data. For this data, we compute the single classifier and combiner classification performances, as well as the Bayes error rate and an error bound.},
	bib2html_pubtype = {Refereed Conference Papers},
	bib2html_rescat = {Bayes Error Estimation},
        year={1995}
}

Generated by bib2html.pl (written by Patrick Riley ) on Tue Jun 26, 2018 19:10:42