Methods for Identifying Systems by Criteria Based on the Minimum Entropy of the Identification Error

The paper presents the description of the problem of identifying the parameters of a system, whose parameterization at a priori and a posteriori stages differs, and the identification criterion is based on involving such an identification error characteristics as the entropy. In this case, the identification criterion is maximized for all possible distributions and then minimized for the vector of system parameters. A heterogeneity measure is then constructed using the Rényi quadratic entropy, which allows the input variables (structural identification as an initial step) to be selected in a suitable way, justified by the properties of the proposed heterogeneity measure, for models subjected to further parametric identification. In this case, the properties of the proposed measure of dependence, namely the consistent normalization of the measure of dependence, are very important.