Emulate QGeNIe's four influence types in GeNIe

[charlie77]

Any advice or reference for this - I want to create the four types of parent-child influences that are neatly offered in QGeNie but not in GeNie - Causation, Absence causation, Requirement, and Inhibition. I want to use GeNIe instead of QGeNie as I have multiple states for a node rather than jus binary.

I tend to think this might be hopeless as QGNie allows this at the arc level so a child node can have different influence types from different parent nodes. However, GeIne only allows changes at the node level - regardless how you manipulate a canonical node it treats all the parents the same way.

The only option, theoretically, seems to be specifying a full CPT in a Chance-General node, but this is way too complicated.

Thanks in advance.

[shooltz[BayesFusion]]

QGeNIe internally builds the CPTs for each node based on the incoming arc types/strengths. You can check it using the "Export to GeNIe" command available in the File menu.

As you pointed out, the algorithm used to populate the CPT is not directly applicable to the scenario where node and/or its parents have more than two state.

[charlie77]

Thanks. Will try Export to GENIE function.
In my case, those types of influence might not be that critical when the subject of interest has passed the early stage of development. The effects of all the variables are becoming more additive.

The main problem I have in GENIe is still the Error(42) UpdateBeliefs failed - network structure is too complex. I have to cut down a lot of details. I don't know whether this is a intrinsic computation constrain with GeNIe or my laptop's limitation (I7, 32G).

[shooltz[BayesFusion]]

I don't know whether this is a intrinsic computation constrain with GeNIe or my laptop's limitation (I7, 32G).

If you can share your model (and the evidence set, preferably as a case(s) defined in .xdsl) we'll be able to check the triangulation statistics using our internal tools.