Setting a Value as Evidence

The front end.
Post Reply
joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

Setting a Value as Evidence

Post by joergweber79 »

Hi all,

suppose we have some nodes with two states (yes/no). Several other BBN tools like hugin or netica have the possibility to set values as evidence, e.g. (set evidence> 80% yes). Unfortunately genie only can set the evidences (yes/no) as far as i can see. Or is there possibility to set values in genie? If not it would be nice, if you'd remeber this in the next release planning.

Lots of thanks for your nice work so far.

Joerg
mark
Posts: 179
Joined: Tue Nov 27, 2007 4:02 pm

Post by mark »

joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

Is such a feature planned?

Post by joergweber79 »

Hello mark,

thanks for your advice, but it seems that there is no pragmatic possibility to set values as evidence. Is such a functionality planned for future releases? I really need to set some values as evidence in my work. It would be great if you could implement it.

Greetings,

Joerg
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: Is such a feature planned?

Post by marek [BayesFusion] »

joergweber79 wrote: thanks for your advice, but it seems that there is no pragmatic possibility to set values as evidence. Is such a functionality planned for future releases? I really need to set some values as evidence in my work. It would be great if you could implement it.
Jeorg,

Yes, we plan to develop this and add to GeNIe. There is a graduate student interested in this topic and he will be working on a thesis in this area. Unfortunately, this means that the estimated release date is a year or so from now :-(. But, the good thing is that it will be done well, like everything in GeNIe and SMILE :-).
Cheers,

Marek
joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

Post by joergweber79 »

Hello Marek,

how is the current situation with setting a value is evidence? Is it still planned to be implemented suddendly? Hopefully your graduate student is still interested in this topic :)

I'm asking because it's really important for me that this feature will be implemented, if not i have to find another solution. This feature is absolutely neccessary for my further work. Could you please give me an estimation, when it will be implemented?

I'm really happy about having Genie & Smile and it would be perfect if you would consider this feature.

Greetings,

Joerg
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Post by marek [BayesFusion] »

joergweber79 wrote:how is the current situation with setting a value is evidence? Is it still planned to be implemented suddendly? Hopefully your graduate student is still interested in this topic :)

I'm asking because it's really important for me that this feature will be implemented, if not i have to find another solution. This feature is absolutely neccessary for my further work. Could you please give me an estimation, when it will be implemented?

I'm really happy about having Genie & Smile and it would be perfect if you would consider this feature.
Dear Jorg,

We are planning to add a basic implementation of "soft evidence" within the next few weeks. Please stay tuned.
Cheers,

Marek
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Feature implemented

Post by shooltz[BayesFusion] »

Soft evidence was added to GeNIe in build 2.0.3568.0.
joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

Thank You so much! But a small remark...

Post by joergweber79 »

Thank you so much, you saved my thesis :)!!! Soft evidence works perfect for me and it is implemented exactly as i expected it to be! Nice work! The only thing i am wondering about is, why the symbol for the soft evidence dissapears if you click on the yellow update arrow (or update immediately via the menu). In a large net with a lot of soft evidences you can't remeber all nodes where soft evidence has been entered. So a visible indicator for an entered evicende would be good. And i'm wondering why i can't delete the soft evidences via "clear all evidences".

Thank you so much again!

Greetings,

Joerg
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: Thank You so much! But a small remark...

Post by shooltz[BayesFusion] »

I can't reproduce the issues you've reported.

1) when I invoke 'Clear All Evidence', the soft evidence is cleared
2) after 'Update Beliefs' the soft evidence nodes keep their soft evidence (the light gray grounding icon remains, and evidence values are still available).

Can you send me your network with the descriptions of the steps required to reproduce the problem?
joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

No evidence sign present

Post by joergweber79 »

I attached an example to this post. And it seems to be that the problem has something todo with the submodel 'B' in the model. If i delete the submodel 'B' everything works as it should.

You can reproduce the problem by the following:

1. Go to submodel A
2. Select an arbitraty node(s) for setting the soft evidence
3. Click on update (yellow arrow)

No icon can be seen within the nodes.

If you have enabled the "automatically update" option, you have the same problem.
Attachments
Network_Problem_Soft_Evidence.xdsl
Problems with setting soft evidence in Submodel A
(25.7 KiB) Downloaded 386 times
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: No evidence sign present

Post by shooltz[BayesFusion] »

joergweber79 wrote:I attached an example to this post. And it seems to be that the problem has something todo with the submodel 'B' in the model. If i delete the submodel 'B' everything works as it should.
It's not the submodel B itself, but rather the presence of the utility nodes. At this point, the inference algorithm for influence diagrams doesn't recognize soft evidence and re-initializes the related data structures associated with SMILE's DSL_node. The end effect is that soft evidence is lost when update is performed.

I'm checking the possible fixes for the issue.
joergweber79
Posts: 21
Joined: Thu Jan 15, 2009 4:22 pm

Post by joergweber79 »

Is a fix or a strategy for fixing already available? Can you estimate when there will be a working version available?
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Post by shooltz[BayesFusion] »

joergweber79 wrote:Is a fix or a strategy for fixing already available? Can you estimate when there will be a working version available?
The preview version of GeNIe with the fix can be downloaded from:
http://genie.sis.pitt.edu/download/genie2_binaries.zip

(this file contains binaries only - you'll need to extract it into the existing GeNIe installation directory).
Post Reply