Change in the bayesian network structure

Question

I'm learning a bayesian network and was wondering if it is possible to merge multiple children into a single child? For example in the figure below, could it be possible to have a single conditional probability table (Node DEF) from the three conditional probability tables (D, E, F).

If it's not possible, is there any work to make independent events to dependent event?

Thank you all

Answer 1

This question is not very well asked, so it is difficult to answer.

First of all the nodes of a BN are not events but random variables.

Secondly, merging nodes requires a justification of this merge. The first idea would be to generate a multidimensional variable with 8 values, which represents exactly the 8 possible values of DEF.

It is also possible to set up merging solutions like aggregators: an AND, an OR, a noisyOR, etc...

Finally, there is a large literature on sensor data fusion that certainly deserves to be looked at closely.