A unified approach to fuzzy pattern recognition

Goldfarb has proposed a unified approach to pattern recognition. In Goldfarb's approach the data from a pseudometric space are isometrically embedded in a pseudo-Euclidean space. The data representation is built for every particular data set. The resulting data space will be data dependent. The aim of this paper is to extend Goldfarb's approach for fuzzy clustering. A fuzzy clustering procedure for a pseudometric data set is given. A generalisation of this algorithm to obtain the cluster substructure of a fuzzy class is also proposed.