Learning to Play Connect 4: A Study in Attribute Definition for ID3
The use of algorithms such as ID3 to induce decision trees and rule sets requires that a set of attributes or features be defined with which to describe objects to be classified. This problem is considered in an application to the game of Connect 4 where the task is to learn a set of rules with which a program can play to a reasonable standard. The attributes used evaluate the current position of a game from the point of view of both players and therefore, to a limited extent, implement a defensive as well as an offensive strategy. The attributes characterise moves made by the ultimate winners in a series of games played by novice and moderately good players.
KeywordsGame Playing Data Collection Form Rule Induction Board Position Attribute Definition
Unable to display preview. Download preview PDF.
- Bratko, I. (1990). Prolog programming for artificial intelligence. Wokingham: Addison-Wesley.Google Scholar
- Epstein, S. L. (1990). Learning plans for competitive domains. In B. W. Porter & R. J. Mooney (Eds.), Proceedings of the seventh international conference on machine learning (pp. 190–197). San Mateo: Morgan Kaufmann.Google Scholar
- Matheus, C. J. (1991). The need for constructive induction. In L. A. Birnbaum, & G. C. Collins (Eds.), Proceedings of the eighth international workshop in machine learning (pp. 173–177). San Mateo: Morgan Kaufmann.Google Scholar
- Quinlan, J.R. (1986). Induction of decision trees. Machine Learning, 1, 81–106.Google Scholar