WEIGHTED PARZEN WINDOWS FOR PATTERN-CLASSIFICATION

Authors
BABICH GA CAMPS OI
Citation
Ga. Babich et Oi. Camps, WEIGHTED PARZEN WINDOWS FOR PATTERN-CLASSIFICATION, IEEE transactions on pattern analysis and machine intelligence, 18(5), 1996, pp. 567-570
Citations number
8
Categorie Soggetti
Computer Sciences","Computer Science Artificial Intelligence","Engineering, Eletrical & Electronic
Journal title
IEEE transactions on pattern analysis and machine intelligence → ACNP
ISSN journal
01628828
Volume
18
Issue
5
Year of publication
1996
Pages
567 - 570
Database
ISI
SICI code
0162-8828(1996)18:5<567:WPWFP>2.0.ZU;2-T
Abstract
This correspondence introduces the weighted-Parzen-window classifier. The proposed technique uses a clustering procedure to find a set of re ference Vectors and weights which are used to approximate the Parzen-w indow (kernel-estimator) classifier. The weighted-Parzen-window classi fier requires less computation and storage than the full Parzen-window classifier. Experimental results showed that significant savings coul d be achieved with only minimal, if any, error rate degradation for sy nthetic and real data sets.