A conceptual framework and a heuristic program for the credit-assignment problem

摘要

The task of credit assignment is interpreted as the ability of a system to 1) identify and distinguish strategy components; 2) associate with such components, in different regions of the domain of confrontation, good and poor outcomes of a sequence of actions prescribed by the strategy; 3) improve the boundaries of these regions in minimizing the errors of misclassification of strategy actions; 4) estimate the overall quality of a strategy being the sum of the qualities of strategy components weighted by the probabilities of employing them; and 5) provide information for a metastrategy that shifts the domain of confrontation to those regions in which the strategy studied is most proficient and thereby raises the effective quality of the strategy. In this correspondence the program QO-4 — the fourth module of a large system, the quasi-optimizer (QA), which can accomplish the above tasks within certain limitations, is described.