[For the entire collection see Zbl 0517.00013.] The aim of this paper is to introduce and to study a new general class of learning algorithms: absorbing barrier algorithms of the reward-penalty type with identical behaviour under the occurrence of success and failure; conditions are obtaining for strong absolute expediency (original concept) and for $ϵ$-optimality of these algorithms. The paper is of interest in mathematical psychology and in learning automata theory and mathematical statistics.
Reviewer: L.Olaru