ON SYSTEM IDENTIFICATION, DECOMPOSITION AND CONSOLIDATION.

The identification, decomposition and consolidation (aggregation) of a dynamic system are studied systematically. The discrimination measure, the organization measure, and the coarseness measure of consolidation are respectively defined as the degrees of nonoptimality of identification, decomposition, and consolidation. These measures are the so-called information measures which are the relative entropy functions, and their meanings and common properties are illustrated. By means of these measures, the optimal system model (state of nature), the optimal decomposition, and the optimal consolidation are defined, and it shown that the optimal ones can be obtained on-line by the Bayesian approach.