A Distributed Clustering Algorithm Prototype Prototype Prototype

A new algorithm for clustering is presented — the Distributed Clustering Algorithm (DCA). It is designed to be incremental and to work in a real-time situation, thus making it suitable for robotics and in models of concept formation. The DCA starts with one cluster (or rather prototype at the center of the cluster), successively adding prototypes and distributing them according to data density until a certain criteria is fulfilled. This criteria is that new prototypes do not add enough extra precision in the representation of the data. A local measure called roundness is used to predict how much extra precision a new prototype will add.