Learning the risk board game with classifier systems
Atila Neves, Osvaldo Brasāo, Agostinho C. Rosa · 2002
The goal is to produce agents that are able to play the board game efficiently. Classifier Systems (CS) were chosen to learn the task at hand. CS were used to learn how to classify a set of (state, action) pairs. These pairs represent a game situation and the action a sensible player should execute when faced with such a situation. Results show that the CS agents perform poorly when compared to humans, but can hold their own in specific situations against computer agents with a fixed, pre-programmed strategy.