Xmas coding!

I just passed Christmas with my family, so when the dinner has finished I just come back to my computer and continue coding my final prototype for my thesis. I've already obtained results but I'm working to improve the process. I've no time to lose because I need it ready for this week. The good news is that I've already a JADE-based Multi-agent system running where each agent is using a cognition module that employs as base a Bayesian program to make decisions.

So, obtain this system cost me a lot of time and effort because from the last year when I took the course until now I've learned by myself how to create and implement these the Bayesian algorithm using Java and merge it with the multi-agent system approach under JADE, a hard task for a newbie in statistics.

Why not use C++? the reason is that there is not a viable solution similar to JADE for Java. However my next years' plan is to build an open source C++ based framework for agents using the Nokia's SDK Qt as basis.

Well returning to the main topic, this Christmas was a debugging and improving experience. My prototype is almost ready. The next step is write the system output in a LaTeX document using the obtained results and improve some parts of the text. Work never ends. When I was a child I never imagine to pass a Christmas working hard as now. My gift list was composed by obtain a working MAS prototype for Christmas! Time changes the people goals.

Ph. D. Thesis Update November 2009

In the last months my thesis work has been evolved drastically. So I found a lot of answers to the main questions that I was looking for and that caught the advance of my thesis work.

Well, now I know that the researcher instinct could not fail since I started with this thesis work in the 2007's summer I said "I need a statistical and probabilistic course" because my engineer formation only give me some hints about these topics. But unfortunately I didn't find a course like that in Cinvestav in my first Ph. D. phase then I tried to get in touch with a professor at Cinvestav that works in the stochastic systems area but I received an "encouraged" answer saying that I must go and search myself these knowledge in the books or literature.



The time passed and I focused in other areas as the Meta-analysis in medical area, theory of emergence behavior, model driven engineering, meta-models, etc. Then a good day in October I received an email from Dominique inviting me to take a course about Bayesian cognition in Grenoble and just I was reading a book that I found with my colleague Anissa in a "secret crate" about the probabilistic reasoning in multi-agent systems. I was taking the ideas from the book as a possible way to solve my thesis issues then I decided to take the course in Grenoble.



Each session was as if all the answers appeared as a slow dropping fall of water, they show me:
- Basic probabilistic notation and operations that are used in the Bayesian cognition area.
- How to treat in an uncertain enviroment (as the problem description phase in my thesis).
- Examples of solutions applied in the industry and scientific problems solved with it.
- How to create a statistical model that could be used in every solution that uses this approach.
- The different kind of induction algorithms to predict most accurate choices.


So, it was a very short course and I spend a lot of money going each week to Grenoble but I think that it was a very good investment. Now I've a clearest idea of how to solve and implement the problem description phase uncertainty issues and then how to manage them into the next phases.

Special thanks to Pr. Pierre Bèssiere and Pr. Julien Diard.