Modeling of human strategy in rule discovery and it's application to multi-agent systems

摘要

We know that our rule discovery speed in model selection problem is very fast comparing with that of machine learning systems. This suggests that the human rule discovery strategy is different from that of common data mining algorithms. When we can revel the human strategy, we can apply it to wide applications. So, in this paper, we analyzed a psychological experiment to model human strategy. Then, we applied the strategy to a multi-agent learning systems, and analyzed its learning dynamics.