Pratical results from artificial economy with application to reinforcement learning

Abstract

This thesis discusses the topic of learning in a complex environment. Both the historical basis of the field and a broad selection of the current works are summarized. Reinforcement learning is the problem faced by an agent that learns behavior through trial-and-error interactions with a dynamic environment. The work described here has a resemblance to work in psychology, but considerably in the details and in the use of the word "reinforcement". We describe the foundations of new field of reinforcement learning named artificial economy. We experiment with a prototype based on the artificial economy model to solve the Blocks World problem and show how the artificial economy approach can be integrated with other reinforcement learning techniques.