PATTERN-BASED LEARNING AND SPATIALLY ORIENTED CONCEPT-FORMATION IN A MULTIAGENT, DECISION-MAKING EXPERT

As they gain expertise in problem solving, people increasingly rely on patterns and spatially oriented reasoning. This paper describes an as sociative visual-pattern classifier and the automated acquisition of n ew, spatially oriented reasoning agents that simulate such behavior. T hey are incorporated into a multi-agent game-learning program whose ar chitecture robustly combines agents with conflicting perspectives. Whe n tested on three games, the visual-pattern classifier learns meaningf ul patterns, and the pattern-based, spatially oriented agents generali zed from these patterns are generally correct. The accuracy of the con tribution of each of the newly created agents to the decision-making p rocess is measured against an expert opponent, and a perceptron-like a lgorithm is used to learn game-specific weights for these agents. Much of the knowledge encapsulated by the new agents was previously inexpr essible in the program's representation and in some cases is not readi ly deducible from the rules.