Hi,
I am wondering how to do structure learning for DBNs, i.e. what the dataset look like and what are the steps to learn the structure using Smilearn/Genie.
I can create a dataset where each record contains value for each variables in two time slices. For example:
a b c a_0 b_0 c_0
1 10 3 2 20 5
...
where a, b and c are in slice 1 and a_0, b_0, c_0 i slice 2.
However, the network structure learned from this dataset by PC looks like a static network with 6 variables (a, b, c, a_0, b_0, v_0).
It would be very useful if anyone could provide an example or explanation on how to do this. Thanks.
Structure learning for DBN
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Structure learning for DBN
Structure learning in SMILE does not support the DBNs. You can learn DBN's parameters with EM.petcai wrote:I am wondering how to do structure learning for DBNs
Re: Structure learning for DBN
Thank you, shooltz. Is there any chance that the documents for smile/smilearn will be updated soon? It looks that documents for quite a number of classes are out of date. Thanks.
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Structure learning for DBN
I don't think we'll be able to allocate resources for documentation in the short term. However, we are constantly monitoring the forum, so feel free to post any questions here.petcai wrote:Thank you, shooltz. Is there any chance that the documents for smile/smilearn will be updated soon? It looks that documents for quite a number of classes are out of date. Thanks.