Using continuous variables about libspn

[jostosh]

Hi Lapayo,

I am very sorry about the late reply. Yes there is support, but it is quite limited in pronobis/libspn. I suggest you have a look at feature/conv_mvg_rat in my fork of the repo. It is still undergoing quite some refactoring, but there are numerous leaf distributions implemented there in libspn/graph/distribution.py (and at least these are tested thoroughly in unit tests and experiments, so you should be able to use these safely). If you have any questions on how to use them, please tell me! The univariate nodes (e.g. CauchyLeaf and NormalLeaf) work similar to IVs: they also have a num_vars parameter, but rather than num_values they have a num_components parameter. Multivariate leaf distributions also have a dimensionality parameter.

I'll try my best to respond sooner from now on!

All the best,
Jos

[Lapayo]

Hi @jostosh ,

thank you very much for your answer! I am very eager to test your fork, but unfortunately I dont have access to it. Would you mind sharing it with me?

Thanks in advance!

Kind regards,
Simon