Nikunj C. Oza's Publications

Sorted by DateClassified by Publication TypeClassified by Research Category

Input Decimated Ensembles: Decorrelation through Dimensionality Reduction

Input Decimated Ensembles: Decorrelation through Dimensionality Reduction. Nikunj C. Oza and Kagan Tumer. In Proceedings of the Second International Workshop on Multiple Classifier Systems, pp. 238–249, Springe-Verlagr, Cambridge, UK, June 2001.

Download

[PDF]181.5kB  

Abstract

Using an ensemble of classifiers instead of a single classifier has been shown to improve generalization performance in many machine learning problems [4, 16]. However, the extent of such improvement depends greatly on the amount of correlation among the errors of the base classifiers [1,14]. As such, reducing those correlations while keeping the base classifiers' performance levels high is a promising research topic. In this paper, we describe input decimation, a method that decouples the base classifiers by training them with different subsets of the input features. In past work [15], we showed the theoretical benefits of input decimation and presented its application to a handful of real data sets. In this paper, we provide a systematic study of input decimation on synthetic data sets and analyze how the interaction between correlation and performance in base classifiers affects ensemble performance.

BibTeX Entry

@inproceedings{oztu01,
        author={Nikunj C. Oza and Kagan Tumer},
        title={Input Decimated Ensembles: Decorrelation through
                Dimensionality Reduction},
        booktitle = {Proceedings of the Second International Workshop on
                Multiple Classifier Systems},
        publisher = {Springe-Verlagr},
	month = {June},
        address={Cambridge, UK},
        editor = {Kittler, J. and Roli, F.},
        pages = {238-249},
abstract={Using an ensemble of classifiers instead of a single classifier has been shown to improve generalization performance in many machine learning problems [4, 16]. However, the extent of such improvement depends greatly on the amount of correlation among the errors of the base classifiers [1,14]. As such, reducing those correlations while keeping the base classifiers' performance levels high is a promising research topic. In this paper, we describe input decimation, a method that decouples the base classifiers by training them with different subsets of the input features. In past work [15], we showed the theoretical benefits of input decimation and presented its application to a handful of real data sets. In this paper, we provide a systematic study of input decimation on synthetic data sets and analyze how the interaction between correlation and performance in base classifiers affects ensemble performance.},
bib2html_pubtype = {Refereed Conference},
bib2html_rescat = {Ensemble Learning},
        year = {2001}
}

Generated by bib2html.pl (written by Patrick Riley ) on Sun Jan 13, 2008 22:02:08