PROBABILITY PROBLEM ABOUT INSTRUCTION OF AUTOMATA TO DISTINGUISH CLASSES AND THE METHOD OF POTENTIAL FUNCTIONS

There is given a probability formulation of the problem of instruction of automata to separate input situations into two classes. An account is given of an algorithm based on the method of potential functions which makes it possible to construct a function which approximates the degree of certainty that the situation arising at the input of the automation belongs to one class or the other. A theorem is proven concerning the convergence (in probability) of the function set up by the algorithm to this degree of certainty. (Author) Edited machine trans. of mono. Veroiatnostnaya Zadacha ob Obuchenii Avtomatov Raspoznavaiyu Klassov i Metod Potentsialnykh Funktsii, Moscow, 1964 p1-40.