Learning Structure

The front end

Learning Structure

Postby borisrabin » Sun Oct 03, 2010 11:46 pm

Hello ,

I went through the tutorial 15:Learning Structure .

I took for example the "retention.txt" data set.

Data -> Learn New Network - > pc (All by the example) but i get some new screen by the name : "Pattern Editor-PC" and when i pressed "Only Create Structure" i get the error "Pattern is not DAG...."

What is that screen and why i can not only get the structure from the data like the tutorial present ?

Another Question:
Supposing i got the Learned the Bayesian Network , can i store the result at Text file for using it later in my code?

Thanks in advance ,

Boris
borisrabin
 
Posts: 24
Joined: Thu Sep 30, 2010 7:48 pm

Re: Learning Structure

Postby shooltz » Mon Oct 11, 2010 11:25 am

borisrabin wrote:Data -> Learn New Network - > pc (All by the example) but i get some new screen by the name : "Pattern Editor-PC" and when i pressed "Only Create Structure" i get the error "Pattern is not DAG...."

What is that screen and why i can not only get the structure from the data like the tutorial present ?


The tutorial was written before we changed our PC implementation. The new implementation creates the structure from data, but the structure is not necessarily a directed acyclic graph (DAG) - it can contain cycles and/or uni- and bi-directional links. This structure must become a DAG before SMILE/GeNIe can use it to create a BN - this is the reason for the new dialog box.


Supposing i got the Learned the Bayesian Network , can i store the result at Text file for using it later in my code?


You can store it using any file format which is supported by SMILE. All of them are text-based.
shooltz
Site Admin
 
Posts: 812
Joined: Mon Nov 26, 2007 5:51 pm

Re: Learning Structure

Postby borisrabin » Mon Oct 11, 2010 2:59 pm

shooltz wrote:
borisrabin wrote:Data -> Learn New Network - > pc (All by the example) but i get some new screen by the name : "Pattern Editor-PC" and when i pressed "Only Create Structure" i get the error "Pattern is not DAG...."

What is that screen and why i can not only get the structure from the data like the tutorial present ?


The tutorial was written before we changed our PC implementation. The new implementation creates the structure from data, but the structure is not necessarily a directed acyclic graph (DAG) - it can contain cycles and/or uni- and bi-directional links. This structure must become a DAG before SMILE/GeNIe can use it to create a BN - this is the reason for the new dialog box.



Supposing i got the Learned the Bayesian Network , can i store the result at Text file for using it later in my code?


You can store it using any file format which is supported by SMILE. All of them are text-based.


* What is the new structure represent?
* How can i build BN from my data (with this new structure representation ), when my main purpose is finding dependency between
attributes?
I attached DB for example.

Thanks,

Boris
borisrabin
 
Posts: 24
Joined: Thu Sep 30, 2010 7:48 pm

Re: Learning Structure

Postby shooltz » Mon Oct 11, 2010 5:24 pm

I attached DB for example.


Your post contains no attachments.
shooltz
Site Admin
 
Posts: 812
Joined: Mon Nov 26, 2007 5:51 pm

Re: Learning Structure

Postby borisrabin » Mon Oct 11, 2010 9:58 pm

shooltz wrote:
I attached DB for example.


Your post contains no attachments.
Attachments
Database1.zip
Data Base
(85.11 KiB) Downloaded 46 times
Database1.rar
Data Base
(70.73 KiB) Downloaded 55 times
borisrabin
 
Posts: 24
Joined: Thu Sep 30, 2010 7:48 pm

Re: Learning Structure

Postby mark » Tue Oct 12, 2010 9:15 am

borisrabin wrote:* What is the new structure represent?
* How can i build BN from my data (with this new structure representation ), when my main purpose is finding dependency between
attributes?

The new structure denotes a direct dependency between variables when an edge exists, and edges are oriented if the algorithm could decide it should be. You can build a BN from this structure by creating a DAG out of it.
mark
Site Admin
 
Posts: 178
Joined: Tue Nov 27, 2007 4:02 pm

Re: Learning Structure

Postby borisrabin » Tue Oct 12, 2010 7:57 pm

mark wrote:
borisrabin wrote:* What is the new structure represent?
* How can i build BN from my data (with this new structure representation ), when my main purpose is finding dependency between
attributes?

The new structure denotes a direct dependency between variables when an edge exists, and edges are oriented if the algorithm could decide it should be. You can build a BN from this structure by creating a DAG out of it.


Thanks for your answer,

My only purpose is to learn Bayesian Network from Database (Attached above) which represent dependency between attributes .
How i can do it directly without getting this intermediate stage ? Which algorithm i should choose ? maybe i need to discrete my variable ?

Thanks in advance ,
Boris
borisrabin
 
Posts: 24
Joined: Thu Sep 30, 2010 7:48 pm

Re: Learning Structure

Postby mark » Tue Oct 12, 2010 9:39 pm

PC is the only option for continuous data, otherwise you need to discretize.
mark
Site Admin
 
Posts: 178
Joined: Tue Nov 27, 2007 4:02 pm


Return to GeNIe

Who is online

Users browsing this forum: No registered users and 1 guest