Preprocessing: Using different number of jets for each sample in the pdf resampling
Currently all samples are up/down sampled to the same number of jets. In order to train on subcategories of light/b/c jets it would be useful to be able to perform the resampling of the subcategories in a way that the total number of light, c and b-jets is still the same. E.g. if b-jets are split in two subcategories the number of jets in subcategory 1 plus the number of jets on subcategory 2 should add up to the number of c-jets/light jets.