[SciPy-user] Bayes net question

David Warde-Farley dwf@cs.toronto....
Mon Jan 14 18:43:45 CST 2008


On 14-Jan-08, at 6:53 PM, Karl Young wrote:

>  From the little I've seen of the package it looks to me like it would
> be worth turing into a scikit and I would be willing to contribute  
> some
> time to such an effort. But it would be nice to get input from any  
> Bayes
> net experts and/or anyone who has used this package re. what they  
> think
> of the code and the approach taken.

On a superficial reading it looks like it might be the beginning of  
something good, though it looks as though it's been abandoned in its  
fairly early stages -- the restriction to discrete variables limits  
it's usefulness in many areas. (Not to criticize Elliot's work, a  
general purpose Bayes net toolbox is a lofty goal and he seems to  
have made significant progress).

I've heard much praise for Kevin Murphy's toolbox for Matlab ( http:// 
www.cs.ubc.ca/~murphyk/Software/BNT/bnt.html ), and this might be  
helpful in extending PBNT or starting another Bayes net Python  
library. Since it is released under the LGPL, any straightforward  
port of Kevin's code to Python would likely need to live in scigpl.

David


More information about the SciPy-user mailing list