Update error

The engine.
lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Update error

Post by lablonsk » Fri Oct 11, 2019 6:33 pm

Hello,
My code worked naturally but when I used a larger data array the following error appeared.
UpdateBeliefs failed, SMILE error code -42
Thanks!

shooltz[BayesFusion]
Site Admin
Posts: 1256
Joined: Mon Nov 26, 2007 5:51 pm

Re: Update error

Post by shooltz[BayesFusion] » Tue Oct 15, 2019 11:44 am

Try marking some nodes as targets. Alternatively, switch to a sampling algorith, like EPIS:

Code: Select all

net.BayesianAlgorithm = Network.BayesianAlgorithmType.EpisSampling;

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Tue Oct 15, 2019 3:11 pm

shooltz[BayesFusion] wrote:
Tue Oct 15, 2019 11:44 am
Try marking some nodes as targets. Alternatively, switch to a sampling algorith, like EPIS:

Code: Select all

net.BayesianAlgorithm = Network.BayesianAlgorithmType.EpisSampling;
The error is saying that it did not find the node because it exists RB. Was it a mistake in learning? Maybe going through matrix I could solve.

"Try marking some nodes as targets." Could you tell me more about this?
I'm using a random try and catch from 0 to 1 but I needed to follow it naturally. I used this command

Code: Select all

 net.BayesianAlgorithm = Network.BayesianAlgorithmType.EpisSampling; 
However the updates were not made and the error persisted saying that the node was not found but when opening the file generated by my algorithm that Smile reads GeNIe efficiently.

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Tue Oct 15, 2019 3:16 pm

Error happens on only one node

shooltz[BayesFusion]
Site Admin
Posts: 1256
Joined: Mon Nov 26, 2007 5:51 pm

Re: Update error

Post by shooltz[BayesFusion] » Tue Oct 15, 2019 5:37 pm

Can you post your network here?

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Fri Oct 18, 2019 6:20 pm

shooltz[BayesFusion] wrote:
Tue Oct 15, 2019 5:37 pm
Can you post your network here?
In private project but generat auto.

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Sat Feb 29, 2020 9:30 pm

shooltz[BayesFusion] wrote:
Tue Oct 15, 2019 5:37 pm
Can you post your network here?
Hello, I'm reopening the topic because I need to create a network using complex information. An exception is returned when performing UpdateBeliefs (), please help. Another thing that is happening is that I needed to process information where the node can be continuous generating an isolated node (this part I am isolating but it is making the process heavy). The most important thing would be to be able to run a complex network. the data have 30 different information but replicated 800 x.
The file is on a website because it is 1.4 MB.
Thanks.
https://drive.google.com/file/d/1cIn8qC ... sp=sharing

shooltz[BayesFusion]
Site Admin
Posts: 1256
Joined: Mon Nov 26, 2007 5:51 pm

Re: Update error

Post by shooltz[BayesFusion] » Wed Mar 04, 2020 11:14 am

Please be more specific - I downloaded your data file. Are you learning both structure and parameters from the data, or you have the network already?

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Wed Mar 04, 2020 8:26 pm

shooltz[BayesFusion] wrote:
Wed Mar 04, 2020 11:14 am
Please be more specific - I downloaded your data file. Are you learning both structure and parameters from the data, or you have the network already?
Hello, I'm learning the structure from this database. When executing the update command an exception is returned.
The maximum number of parents is 3. A posteriori and priori are 0.01 and 0.1 respectively (this is not a rule) but the initial network must be very "messy" because it is an evolutionary process that generates several networks, each network represents a generation .
Last edited by lablonsk on Wed Mar 04, 2020 8:30 pm, edited 1 time in total.

shooltz[BayesFusion]
Site Admin
Posts: 1256
Joined: Mon Nov 26, 2007 5:51 pm

Re: Update error

Post by shooltz[BayesFusion] » Wed Mar 04, 2020 8:28 pm

Which structure learning algorithm is used? If possible, please post your network (after learning, but before the update) here.

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Wed Mar 04, 2020 8:33 pm

shooltz[BayesFusion] wrote:
Wed Mar 04, 2020 8:28 pm
Which structure learning algorithm is used? If possible, please post your network (after learning, but before the update) here.
Attached image of how the process takes place.
Attachments
boa.png
boa.png (53.99 KiB) Viewed 4646 times

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Wed Mar 04, 2020 8:35 pm

shooltz[BayesFusion] wrote:
Wed Mar 04, 2020 8:28 pm
Which structure learning algorithm is used? If possible, please post your network (after learning, but before the update) here.
As soon as I can I send the network to have a look. But if you use the standard that is in GeNIe the software cannot execute the update command because it is very complex.

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Wed Apr 22, 2020 7:31 pm

shooltz[BayesFusion] wrote:
Wed Mar 04, 2020 8:28 pm
Which structure learning algorithm is used? If possible, please post your network (after learning, but before the update) here.
Good afternoon,
If you can take a look I appreciate it. My research depends a lot on me being able to run these RB.

Thank you very much and sorry for the work.
Attachments
dados_IEEE.xdsl
(198.54 KiB) Downloaded 70 times

shooltz[BayesFusion]
Site Admin
Posts: 1256
Joined: Mon Nov 26, 2007 5:51 pm

Re: Update error

Post by shooltz[BayesFusion] » Mon Apr 27, 2020 2:46 pm

Your network is densely connected. Consider reducing the link probability and prior link probability - you have used 0.5 for both. The defaults in GeNIe are 0.1 for link probability and 0.001 for prior link probability.

lablonsk
Posts: 32
Joined: Sat Jun 29, 2019 12:13 pm

Re: Update error

Post by lablonsk » Mon Apr 27, 2020 3:22 pm

Hello, well, the problem is as follows. The problem is that the algorithm is evolutionary, so this same configuration at some point ends up giving some problem. Is there no way around this problem without using a try?
Thank you!

Post Reply