Learning Influence Diagrams from Data

[GimmemoreCoffee]

Hello,

I have got the following question:

Is it possible to learn complete Influence Diagrams with SMILE? It would be nice not only to be able to learn the paramteres of a Bayes Network but also to learn the respective values of a/the utility node/s.

The idea is to use a data-set including the values of the chance nodes and the users decisions, that reflect the maximized utility of the user or something close to it, and then apply an algorithm to learn, both the paramters of the chance nodes and the parameters in the utility node.

Thank you for your help!

[GimmemoreCoffee]

Maybe there is another way to solve the problem. Perhaps I could transform the decision network/ influence diagram into a simple Bayesian Network and apply a standard learning algorithm.
Although it would be much more comfortable to be able to do this with a decision network/ influence diagram.

I don't know why this topic somehow is ignored by everyone.

I would really appreciate an answer, thanks!

[marek [BayesFusion]]

This is a great topic for research. As far as I know, there has been work on this or related topics but all done outside of our lab, so we can't help much. Please start your search from the proceedings of the Annual Conference on Uncertainty in Artificial Intelligence, http://uai.sis.pitt.edu/. I hope this helps.
Cheers,

Marek