Realization of a Stochastic Model by Automata
Ye. S. Usachev · Journal of Cybernetics · 1972
In this paper, which continues [1] and [2], we shall construct a sequence of probabilistic automata that approximate a stochastic learning model. The construction is based on the replacement of a continual automaton (the learning model) by a finite automaton. That such a replacement is possible (and the estimate of the error involved) follows from Theorems 3.1 and 4.1 of [2], which also gives the definition of a perceptron, of a stochastic learning model (BM) and of a particular case of this model (LBM), as well as the proofs of all the theorems cited below.