A FUZZY GENERALIZED NEAREST PROTOTYPE CLASSIFIER
Ludmila Ilieva Kuncheva, James C. Bezdek · 1997
We propose a Fuzzy Generalized Nearest Prototype Classifier (FGNPC). The classification decision is crisp and is based on aggregation of similarities between the unlabeled object x and n p prototypes fv i g with "soft" labels. FGNPC contains as special cases the 1-nearest neighbor rule, the minimum-distance classifier, and some types of radial-basis function networks and fuzzy if-then systems. An experimental illustration is also presented. 1 Introduction The nearest prototype classifier (NPC) is one of the simplest and most intuitively pleasing pattern classification paradigms [3, 6]. Let L = f1; 2; : : : ; cg be a set of class labels and V = fv 1 ; : : : ; vnp g be a set of prototypes, v i 2 ! d ; 1 i n p . We call any function D : ! d ! L a crisp classifier. The classical NPC [6] assumes that n p c, and the prototypes V are crisply labeled to the classes, i.e., a set of crisp class labels I V = fl 1 ; : : : ; l np g; l i 2 L; is associated with V . A vector x 2 ! d ...