Inference of certain networks fails (Exception java.lang.StackOverflowErrorfired in intervention)

[Miguel Ángel Artaso Landa]

The inference of certain networks, like the attached one, is failing. When making click in the inference button, the GUI moves to 'inference mode' but no information regarding the probabilities or the utilities is shown.

The error in the messages window is:

Exception in thread "AWT-EventQueue-0" java.lang.StackOverflowError at org.openmarkov.core.model.network.potential.Intervention.concatenate(Intervention.java:370) org.openmarkov.core.model.network.potential.Intervention.concatenate(Intervention.java:376) [...]

[Manuel Luque]

Please, Miguel, could you reproduce the issue with a smaller influence diagram?

[Miguel Ángel Artaso Landa]

I have been able to eliminate just a few nodes in the network previously attached, so there is not a much simpler network to work with. I am still trying to debug the error to find which is the error. I will keep you posted.

[Manuel Luque]

And, why did you decide to use that network? Have you understood in Jensen's book the particular topology of that ID? Most of the issues can be reproduced with networks of not more than five nodes. So, I first would try to understand on a paper what makes special to that ID. Remember that Jensen and Nielsen have worked a lot about redundant variables in IDs. Debugging is a mess if we don't understand the special problems of that ID. But first I would think if the priority of solving that kind of network is critical.

[Miguel Ángel Artaso Landa]

Well, I was just testing the GUI with several networks when I solved issue #191. The inference with this network failed and I was trying to find what is special about this network. So far, it is the only network I have found that triggers this error. My decision of marking it as critical was solely based in the fact that the inference was not working on it. If it is worth moving to another issue because is not so critical, just let me know.

[Manuel Luque]

I read at the beginning of this issue that you said that the inference with certain networks was failing, but now you say that it is only with that ID? And if issue 191 is more related with the GUI, does it make sense then to test that issue with more complex networks, whose complexity affects more to the inference algorithms than to the GUI? I would use a different test approach... In any case, I reassigned me this issue, because I want to save your time, Miguel.