As a rule, it is possible to construct an adequate probabilistic model (W, ?, P(?))for a statistically stable experiment (control system) Э only in simple cases. To solve this problem for the experiment Э of a complex structure, an approach has been suggested in [1] which is based on a parametric set U with elements u and a family {(W, ?, P
u (?)): u ? U } of probabilistic models. A difficult optimization problem arises here: to choose by some criterion from the family {(W, ?, P
u (?)): u ? U } in some cases, an adequate model for the experiment Э, and, in other cases, a target model for this experiment [2?-4]. Using notation and results from [1], the article considers the problem for a class of permutation and quasi-permutation strategies.
|
