Eventual upgrade of RFPImp version

[GilesStrong]

RFPImp is relied on quite a bit by LUMIN, however it imports stuff from sklearn.ensemble.forest, which is depreciated since 0.22, and will be removed in 0.24. Currently this raises FutureWarning. Hopefully in the next release this will be fixed and we can upgrade LUMIN to use the new version of RFPImp.

[GilesStrong]

Latest release of rfpimp only checks if sklearn version is 0.22 when performing compatibility checks. Results in crashes when trying to use auto_filter_on_mutual_dependence due to _generate_unsampled_indices requiring 3 arguments in sklearn >= 0.22.

Quick fix is to replace if sklearn.__version__.startswith("0.22"): with if LooseVersion(sklearn.__version__) >= LooseVersion("0.22"): and add from distutils.version import LooseVersion to the imports.

[GilesStrong]

Spotted that scikit-learn==0.22.2 meets all the requirements for LUMIN and rfpimp==1.3.5 successfully uses appropriate scikit-learn private API. Adjusted requirements to these versions.