Question about root TMVA

TMVA
1

For a classifier, I want to put some variables in and train them first , then put some other variables and train, in TMVA ,is there any methods to help to calculate the “two-step” training signal and background effieicency.

2

Hi,

Thanks for the post.
Adding in the loop @moneta .

Cheers,
D

3

Hi,
You can do this in TMVA, you create a TMVA::DataLoader class, add first dome variables train and then evaluate the result. Then you create a new DataLoader with the new variables and you trin again and re-evaluate. An example of this is in this code in VariableImportance::EvaluateAll.

Lorenzo

1 Like
4

Thanks , it helps a lot :slight_smile: