A clustering algorithm for vector valued processes

Using a metric criteria, and with no prior information, the author determines the existence of clusters in the set X. He also determines the number of clusters, the cluster membership of each point of X, and consequently a disjoint decomposition of X. He considers whether, if X is a moving window on a possibly infinite data stream, these matters can be resolved online. A methodology is presented for solving the issues posed above. The design has features which are compatible with VLSI technology and, in particular, systolic array implementation. The performance of the design is partially evaluated using a simulation testbed