Computer learning through association
Robert W. Snelsire, R.P. Roesser · NCSU Libraries Repository (North Carolina State University Libraries) · 1968
In this report a computer program is described which attempts to model the techniques by which human beings take information about an initially unknown environment and use it to develop rules for coping with the environment.The environment chosen is a simple game, similar to tic-tac-toe. The computer program and a randomly playing opponent alternately place their symbols on a three by three board. After nine such moves the program is told whether it won or lost the game according to a predetermined set of rules. The model is programmed to develop a strategy for successive games. As the computer plays more games the rules and strategy become more sophisticated.Two fundamental hypotheses were used in the implementation of the learning process.1) Learning is achieved through association on simple observable phenomena (game results), to more complex concepts. This gives direction to the learning process.2) Learning is based on a type of economic system in which concepts are developed only when their worth in playing an effective game exceeds their cost in complexity.The program was written in IPL-V and implemented on the IBM 7044 at the State University of New York at Buffalo.