Identification of distinctive features for classification in complex systems

Abstract

This paper deals with the problem of classifying patterns encountered in complex systems. It describes an approach to pattern recognition that results in a complete and reliable classification technique. It is noted that the majority of existing pattern recognition methods initiate their classification acts on identification of similarities between the members of various classes. On contrast, the work reported here, starts with the recognition of distinctive features of encountered patterns. It is proposed that the patterns to be clustered in a particular fashion to facilitate the exploration of their distinctive features. The process does not depend on utilization of heuristic rules. The membership of different classes will then be based on different values for some or all of such features. This paper will also establish that by utilizing the distinctive features complete classification of all patterns, even for complex systems, can be achieved.