The goal of this page is to be the central point of people interested in working on a toolbox for BayesNetwork?. I (cdavid) got the explicit authorization from Kevin Murphy to port his toolbox in python under a BSD license:
http://www.cs.ubc.ca/~murphyk/Software/BNT/bnt.html
BNT depends on netlab, also available under the BSD.
People interesting in working on the project
- David Cournapeau
- Andrew Straw
- David Warde-Farley
- Karl Young
- Johann Cohen-Tanugi
Plan of action
- Set milestones, approximate timelines?
- Perhaps we should focus on discrete cases first, or something like that, so that we can get up to speed with other Python implementations
- I don't think BNT is very object-oriented, so we will probably need to do some planning as to how exactly to map this to Python while remaining as close to the original as possible, for the sake of switchers (and to make porting easier)
- Divide up responsibilities, so there's no duplication of effort?
Other notes
- See scipy ticket #52, which contains a patch that gets openbayes to run on numpy. Openbayes is inspired by BNT. It works nicely for discrete nodes, continuous don't work in all situations.
