Jason Briceno - 2008-04-03

I'm currently working on a project that would utilize the PNL in a commercial product. In particular, we will be using the BBN and DBN functionality of the library. However, for this project, we will need to implement non-linear, non-gaussian CPDs for some of the nodes. I was hoping to be able to just write an extension for the library, so that I wouldn't have to mess with the existing PNL codebase. However, after becoming a little more familiar with how the PNL works, I don't think that's going to be doable. Ultimately, I'm going to have to make modifications to the sourceforge code in order to add the functionality that I need.


My hope is that I can add this functionality without breaking any of the current functionality or interfaces. Any advice/recommendations would be welcome.