Probabilistic Model Parameter Optimization for the Problem Solving Algorithm

Abstract

The variety of problem solving algorithms models over set of the alternative solutions determines the application of the principle of equivocation logical reduction by narrowing of the solutions set. The choice of an optimum decision comes to the logical conversion of alternative solutions set to the feasible solutions set and to the effective solutions set. The alternative solution set is transformed to the feasible solution on the constraints set. The article explains the application of Hidden Markov Model (HMM) for the choice of optimum problem solving algorithm concerning the observable consistency. In this case we use the maximum likelihood criterion with the constraints in the form of normalizing conditions and semantic measure of the information expedience of A.A. Harkevich for the optimization of unknown parameters of the problem solving algorithm. The “committee” constructions are used for the “integration” of some algorithms for collective decision. We receive the optimal parameters for the algorithm of the collective decision using estimation of the posterior probabilities of algorithm appliance.