TMVA: How to get event-by-event weights?


I am currently working with MC data that doesn’t seem to accurately reproduce the underlying correlations between variables (1D projections look OK), and therefore when I compare the MVA output from my BDT’s between MC and real data there is a substantial discrepancy.
I am now hoping that I can find some event-by-event weights that would rectify this. I am currently looking to follow the steps given in this paper (Reweighting with BDT’s: But, is it possible to find these weights within TMVA? Or do I need to migrate to python(e.g. hep_ml package)? Any help would be greatly appreciated! :smiley:


Hi Aaron,

This capability of reweighing the distribution using BDT is not yet implemented in TMVA. It is something we could maybe implement in the future

Best Regards