EVOLUTION OF FINITE AUTOMATA FOR PREDICTION.
Final rept. 5 Jan 65-5 Jan 66,
GENERAL DYNAMICS SAN DIEGO CA CONVAIR DIV
Pagination or Media Count:
Evolutionary programming is considered from the point of view of various representations of the evolving organism. The finite state machine is shown ro offer an appropriate inheritance. Its memory limitation can be overcome through a modified representation. A series of experiments were conducted to demonstrate that the evolution of such finite state machines of extended memory does indeed provide increased predictive capability against periodic environments which suddenly change their character and environments which require the recognition of higher-ordered Markov processes for their adequate prediction. A mathematical technique was devised for effectively deriving a machine representation for the logic underlying the environment in terms of the logic of the evolved predictor machine. Author