Associative Clustering for Clusters of Arbitrary Distribution Shapes

摘要

A novel neural network, named Associative Clustering Neural Network (ACNN), is developed for clustering data whose underlying distribution shapes are arbitrary. ACNN is a dynamic model that collectively measures and updates the similarity of any two patterns through the interaction of a group of patterns. Such a new measure of similarity helps to achieve more robust clustering performance than using the existing measures that are staticly and individually based on the distances among the isolated pairwise data. The efficience of ACNN has been verified through the performance study.