Prediction and Control through the Use of Automata and Their Evolution

Abstract

Finite-state machines provide a natural means for representing the logic which may underlie a sequence of data derived from a sensed environment or for depicting the transduction between stimulus and response of such an environment. Such representation permits expansion of the logic in terms of arbitrary input and output languages so long as these are expressed within finite alphabets. Further, the machines may be of arbitrary specificity so long as they have only a finite number of states. Thus, no unnatural constraint is imposed, as is so often the case when a sequence of data is expressed in terms of a best linear fit or when a transduction is expressed in terms of a linear difference or differential equation. More powerful logical entities can be considered, but these may be identified with problems which reduce their acceptability. The concept of evolutionary programming was conceived as a means to find a most appropriate finite-state machine for the purpose of prediction or modeling in terms of the available data base and an arbitrary expression of goal (a payoff or error-cost matrix).

Keywords

This publication has 0 references indexed in Scilit: