## How to calculate node states in case of virtual evidence?

The front end.
nirzhor26
Posts: 6
Joined: Thu Jan 19, 2017 7:20 pm

### How to calculate node states in case of virtual evidence?

Can anyone suggest or show me any calculation how nodes are updated in case of virtual evidence given to any child node? Suppose, B is the child node of A. P(A)=0.9., P(B/A)= 0.6 and P(B/A')=0.8. In normal updating, P(B)= 0.62. If I put P(B)=0.9 as virtual evidence, after updating P(A) becomes 0.88 and P(B) becomes 0.94. I want to know how GeNIe is doing this calculation. Any suggestion please?

nirzhor26
Posts: 6
Joined: Thu Jan 19, 2017 7:20 pm

### Re: How to calculate node states in case of virtual evidence?

I am attaching the sample network here.
Attachments
sample network.xdsl

nirzhor26
Posts: 6
Joined: Thu Jan 19, 2017 7:20 pm

### Re: How to calculate node states in case of virtual evidence?

I have solved it. Thanks all.

viliop
Posts: 8
Joined: Mon Apr 23, 2018 6:28 am

### Re: How to calculate node states in case of virtual evidence?

Could you help about the way Genie calculate it?

marek [BayesFusion]
Posts: 270
Joined: Tue Dec 11, 2007 4:24 pm

### Re: How to calculate node states in case of virtual evidence?

Virtual evidence implemented in GeNIe/SMILE is a standard calculation that you can read about in the literature. Here is one place where you can find information: https://www2.ee.washington.edu/techsite ... 4-0016.pdf. Essentially, virtual evidence corresponds to a child node of the evidence node that sends information up to the node in question. Parameters (conditional probability distribution) of the child node conditional on the evidence node correspond to the virtual evidence entered. I hope this helps,

Marek