After quite a period of frustration, I finally managed to create a simple frame with Swing that would be my main agent's interface. I had to resort to a nice and small open-sourced GUI builder called "Abeille Forms Designer" (see image below). It is really cute, and I must say looks really nice. It took me a while to understand and get used to, though. This is because it is based on a popular, open source layout manager for Java called the "JGoodies FormLayout" system. Anyway, it did the job and I successfully created the Bayesian learning agent that uses the genetic search implementation in WEKA to learn the Bayesian network structure from datasets spread over a variety of hosts.

The only problem I'm facing now is displaying the resulting network using the GraphVisualizer that comes with WEKA. Everytime I try to create a frame containing the GraphVisualizer from within the Tahiti server I get a strange NullPointerException. After searching the Web for quite a while, I found a post on the WEKA mailing list mentioning the same problem. It is suggested by the poster that the JVM's version might be the problem. However, I'm not quite sure about that because when I attempt to open the GraphVisualizer from the command line it works fine! Guess I have to go back to WEKA's source code and search for the cause.

One thing that I discovered while surfing the Web is a framework created by IBM for building intelligent agents using machine learning and reasoning called "ABLE" (Agent Building and Learning Environment). I discovered it very late though and I decided to continue on the same track and use WEKA. However, even though mobile agents where never mentioned in ABLE's site at alphaWorks, I wonder if it would be possible to use this framework to build intelligent mobile agents that learn and reason. In the mean time, here is how the main agent's frame finally looks like.