Four approaches are given in [1, 2] to construct probabilistic models for controlling systems operating under conditions of structural uncertainty. The first approach is based on the Kolmogorov model for a controlled experiment from the «black-box» point of view. The second approach concerns the application of conditional probability distributions and the study of corresponding sample spaces for experimental result measurers. The third approach assumes representing the system in the form of its constituting elements, each element being a simpler experiment to be studied using the first two approaches. The disadvantages of these three approaches are mentioned in [1, 2]. In this connection a cybernetic approach is proposed in [1–3] which is based on highlighting general properties of the controlling system of any nature, i.e. its scheme, information, coordinates and function. Using this approach, the problem of choosing an adequate model is solved for the experiment with a permutation family of probabilistic models.
|
