Learning in Games

Java is not working. This has been tested in Firefox, and it works there!

This is the learner described in Section 10.4.2. It implements the learning to coordinate algorithm of Figure 10.9 and described in Examples 10.17 and 10.18.

This implements a simplified (single state) version of the PHC and/or the WoLF-PHC algorithm of Michael Bowling, Manuela Veloso. Multiagent learning using a variable learning rate Artificial Intelligence, Volume 136, Issue 2, April 2002, Pages 215-250. It allows for both PHC and WoLF-PHC agents.

The parameters can be changed. To have a PHC agent, the easiest way is to make deltaWin and deltaLose the same, athough the Java code has a simpler agent that just does PHC.

You can get the code as mas.zip. The GUI is MASLearningGUI.java. The learning agent is LearningAgent.java. The WoLF learning agent is WolfLearningAgent.java. The initial game is in SampleGame.java or SampleSoccerGame.java or SampleCoordGame.java. You can also get the javadoc.

This applet comes with ABSOLUTELY NO WARRANTY. This is free software, and you are welcome to redistribute it under certain conditions, see the code for more details.

Copyright © David Poole, 2008.