In this report we are going to analyze how an agent with a Bayesian Network can use it for predicting another agent, in order to enhance its performance.
In order to do this we first implement a simple example game which will be used as a basis for our tests. This includes the implementation of two simple agents, which can be used in conjunction with the more advanced AI.
We are going to model the Bayesian Networks needed for predicting the agents of the game, followed by the training of the Bayesian Networks and tests to measure how well they perform. We found that one of the models had a problem learning the strategy of the agent it was supposed to learn from. Reasons for this were found and discussed.
We are also going to discuss our experiences with Bayesian Networks through this project, and compare it to some of the other technologies which could have been used, as well as come with some propositions for future work.
In the conclusion we found that the approach we have used had some problems. The main problem was the size of the conditional probability tables, which prevented the agent from learning effectively.