Re: Learning Bayesian Networks
Finn V. Jensen (fvj@cs.auc.dk)
Wed, 29 Jul 1998 16:15:01 +0200
Jie Cheng wrote:
>
> The BN PowerConstructor can start learning with an intermediate model --
> just input the edges as domain knowledge in Step 4. This will speed up the
> learning process.
> However, the program may delete some of those edges later on if it finds
> that they are not necessary.
> Maybe I should add an option to the program to force it accept those
> edges.
> The program can be downloaded from http://193.61.148.131/jcheng/bnpc.htm
> or http://infosys.susqu.edu/bnpc/.
>
> Regards,
> Jie Cheng
>
> On Wed, 22 Jul 1998, Ursula Robers wrote:
>
> > Dear colleagues,
> >
> > does anyone know a tool for learning the structure of Bayesian Networks
> > with the ability to start learning with an intermediate model, that
> > means the network may already contain some (sure) edges when learning
> > starts.
> >
> > Thanks in advance,
> > Uschi Robers
> >
I would not consider "sure" edges as particularly usefull information
for a structural learning system. I would guess that "sure" non-edges -
i.e. "sure" conditional independences would have much greater importance
on peformance.
/Finn V. Jensen