A fuzzy model for learning in automata

Publisher Website
Google Scholar
Abstract
SUMMARY A simple general model for learning, using a fuzzy set theoretic approach and fuzzy decision in an automaton which has nonfuzzy input/output, is proposed. The process has been modelled somewhat in the fashion of general biological systems, which may be viewed as a fuzzy decision process where learning consists in taking a tentative action and reinforcing the membership values on the basis of the results of that action. The model is tested on an automaton whose sole purpose is to follow the boundary on an object with which it makes contact during its movements. The automaton is simulated by a computer. it has standard 8–neighbourhood configuration with binary sense capability and three action capabilities. The automaton has been found to learn to take correct action in a large number of possible input situations within only a few thousand moves.

This publication has 1 reference indexed in Scilit: