Switch RVM support to EMRVC

[halsw]

Hello, and congratulations for your work. I've used random forest and SVC on complex datasets and work like a charm, on the other hand I didn't manage to get either RVC(sklearn_bayes) working properly or port_rvm, so my suggestion is to switch RVM support to EMRVC(sklearn_rvm)

sklearn_bayes fails to install, probably because it supports an older version of sklearn. I've managed to install it by modifying the code to support sklearn 0.24 and manually compiling cython source files, but even then it does not work properly and the C++ models I've managed to produce all seem to have 8 relevance vectors. On the flip side sklearn_rvm works really well and it will make your work less sklearn version dependent.

[eloquentarduino]

Hi @halsw, I had troubles installing sklearn_bayes after a while. The reason I chose it over sklearn_rvm was that it was much faster to train. But if it doesn't play well with sklearn, I will work on adding sklearn_rvm support too, so one can choose either of the two.

I marked this issue as "enhancement" for the library.