RFE: OOBN extension of Genie/SMILE?

The front end.
Post Reply
a.s
Posts: 10
Joined: Thu May 10, 2012 10:44 pm

RFE: OOBN extension of Genie/SMILE?

Post by a.s »

Hi,
I am modeling a set of individuals interacting and I must manipulate their individual reactions instead than modeling them as a population represented as a probability distribution.
The obvious way to do it would be to define a subnet as a class and then to instantiate several copies of it.
This would allow the definition of "class CPTs", reducing greatly the number of CPTs to learn from data.
Other bayesian systems (I think Hugin and UnBBayes) allow this kind of OOBN (Object Oriented Bayesian Networks).

I have found a way to emulate a "subnet class" in Genie by using a temporal bayesian network,
and by considering each time slot as if it were one of the class instances.
The machinery works, but forces me to define only regular interconnections among the emulated "instances"
i.e. the individual represented by time-slot X depends on the ones at time-slots X-1, X-2, X-3.

This suggests me that SMILE/Genie has already almost all the machinery required to run OOBN and to learn their class CPTs

It would be wonderful if an OOBN extension of Genie/SMILE would be implemented ... have you already considered the idea?

Thanks
Andrea
Last edited by a.s on Thu May 17, 2012 1:53 pm, edited 1 time in total.
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: RFE: OOBN extension of Genie/SMILE?

Post by shooltz[BayesFusion] »

This suggests me that SMILE/Genie has already almost all the machinery required to run OOBN and to learn their class CPTs
It would be wonderful if an OOBN extension of Genie/SMILE would be implemented ... have you already considered the idea?
We had a brief discussion re: OOBN, but decided not to implement it due to workload required to implement the feature.
Post Reply