Finish mva/tmva_train.py

[cdeil]

@ndawe Did you get a chance to finish the mva/tmva_train.py example? Do TMVA and scikit-learn give consistent (identical?) results?

[cdeil]

I think the scikit-learn example only uses one CPU core? Shouldn't it be using all cores and thus be faster than TMVA?

[ndawe]

scikit-learn offers parallelism when fitting multiple classifiers, like when performing a grid search over classifier parameters and/or cross-validation.

[ndawe]

This scikit-learn vs TMVA example is not yet done. We just need to plot the decision surfaces, score distributions, and ROCs for both.

[ndawe]

The decision tree node splitting could potentially be done in parallel, and I think there has been talk about implementing this in scikit-learn, but at the moment each fitting of a classifiers is a single process.