logistic-cpd nodes

The front end.
Post Reply
liid
Posts: 32
Joined: Tue Jun 19, 2018 5:19 am

logistic-cpd nodes

Post by liid »

Hi,

First, thank you for answering my previous questions!

1. I want to combine one node with a multi-valued logistic cpd (sigmoid) in my network.

I did see that you have the option of an "equation" node but I am not sure how do I need to use it to achieve a multi-valued logistic cpd. For example-I want to learn the weights of the cpd via training and not just determine them as constants in the equation.

(I know that I will probably be able to do that in the code, I am asking if there is a way to do that in the GUI).

2. When I open the "parameter learning" -> learning -> "match network and data" window, there is a yellow exclamation mark near some of the nodes in my network, both in the "network" section and "matches" section.
What does it mean?

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

Re: logistic-cpd nodes

Post by marek [BayesFusion] »

Let me try to answer your questions.

Ad 1: You will need to learn the parameters of your equations as well as possible parameters of continuous noise distributions outside of GeNIe. Once you know what to model, you can enter the equations into your GeNIe model.

Ad 2: The yellow exclamation mark gives you a warning that something is wrong. Quite possibly there is a mismatch between your data and your model. The mismatch may be because of differences in variable or state names or because some of the parents of the nodes are missing.

I hope this helps,

Marek
liid
Posts: 32
Joined: Tue Jun 19, 2018 5:19 am

Re: logistic-cpd nodes

Post by liid »

Hi-

Regarding 2.: I do have missing values in the training data for some of the nodes. I also have latent nodes for which I do not have any training data at all. I this a problem? I thought that since you use EM it should not be..
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: logistic-cpd nodes

Post by marek [BayesFusion] »

GeNIe does not have a function for learning equations or parameters of continuous distributions as of yet (please stay tuned), so you will need to do learning using your own program. I hope this helps,

Marek
Post Reply