No real documentation on creating continuous Bayesian networks

[bramson]

I've read through the documentation and notebooks for Continuous Factors several times, but there is still a big gap between what is written there and what I need to know to create an actual Bayesian network using continuous factors.

It would be great (and I would say essential if you want this library to be used by people) to provide working examples of creating a actual simple Bayesian network including continuous factors in the documentation and/or notebooks. But not overly simple examples so that they can't be extrapolated to real problems.

Your explanations and examples of setting up the conditional probability tables were clear and sufficient. So, essentially, just do what was done for the discrete cases, but for the continuous factors.

Specifically, I can't figure out some basic setup operations. If I have a node C with two continuous inputs A and B, how do I set the parameters of the distribution of C as a function of the distributions A and B. Perhaps as a step in explaining this, you can first explain how to setup the continuous node C where A and B are discrete, then expand that to cover the case where A is discrete and B is continuous, then where they are both continuous.

Also, the LaTeX equations in the documents don't show correctly, they are filled with "unknown character" errors.

[ankurankan]

@bramson Thanks for the feedback. The continuous variable documentation is not clear because the continuous variable features are not yet fully integrated in the package. I have created an issue to add a list of supported features in the documentation so that it's not confusing for the users.

[giri-fiu]

Have the continuous variable features been integrated into the "pgmpy" package? If so, is there new documentation? If not, are there alternative packages you can suggest?

[ankurankan]

@giri-fiu There is partial support (structure learning and some causal inference features) for continuous variables in pgmpy now: pgmpy.org. Continuous Factors are still not supported. Depending on what features you are looking for, you can check out https://github.com/jmschrei/pomegranate.