A merge-based condensing strategy for multiple prototype classifiers

R.A. Mollineda1, F.J. Ferri2, E. Vidal
1Institut Tecnològic dInformàtica, Universitat Politécnica de Valéncia, Valencia, Spain
2Departamento dInformática, Universitat de València, Burjassot, Spain

Tóm tắt

A class-conditional hierarchical clustering framework has been used to generalize and improve previously proposed condensing schemes to obtain multiple prototype classifiers. The proposed method conveniently uses geometric properties and clusters to efficiently obtain reduced sets of prototypes that accurately represent the data while significantly keeping its discriminating power. The benefits of the proposed approach are empirically assessed with regard to other previously proposed algorithms which are similar in their foundations. Other well-known multiple prototype classifiers have also been taken into account in the comparison.

Từ khóa

#Prototypes #Clustering algorithms #Nearest neighbor searches #Neural networks #Adaptive algorithm

Tài liệu tham khảo

kohonen, 1995, Self-Organizing Maps, 10.1007/978-3-642-97610-0 10.1016/0031-3203(93)90040-4 10.1016/0167-8655(94)00070-J 10.1093/comjnl/26.4.354 10.1109/IVELEC.2009.5193407 10.1109/ICNN.1988.23829 king, 1992, Statlog Databases 10.1016/S0167-8655(97)00035-4 10.1109/TSMC.1972.4309137 1997, LVQ PAK The learning vector quantization package 10.1109/21.278999 10.1109/TIT.1968.1054155 chang, 1974, finding prototypes for nearest neighbor classifiers, IEEE Trans Computes, c 23, 1179, 10.1109/T-C.1974.223827 10.1109/5326.661099 10.1109/5326.661091 10.1109/5.58325 devijver, 1982, Pattern Recognition a Statistical Approach mollineda, 2001, Hierarchical agglomerative clustering for prototype selection and distance-based classification use in cyclic strings dasarathy, 1991, Nearest Neighbor (NN) Norms NN Pattern Classification Techniques webb, 1999, Statistical Pattern Recognition