search

arogozhnikov / hep_ml

Machine Learning for High Energy Physics.
https://arogozhnikov.github.io/hep_ml/
Other
176 stars 64 forks source link

Odd behaviour of GBReweighter #59

Open marthaisabelhilton opened 5 years ago

marthaisabelhilton commented 5 years ago

I am trying to use GBReweighter and am getting odd behaviour of the weights as seen in the figures attached. My parameters are as follows: reweighter = GBReweighter(n_estimators=40, learning_rate=0.1, max_depth=3, min_samples_leaf=1000, gb_args={'subsample': 0.4}) I have tried varying the parameters. Do you know what might cause this behaviour? Many thanks. D02KSPiPiDD_2012_original.pdf D02KSPiPiDD_2012_reweighted.pdf

arogozhnikov commented 5 years ago

Can it be so that you have 1) too few samples in MC and 1000 elements in the leaf are almost completely coming from real data? 2) or there are regions in variable space where you have almost no MC? (I blindly guess that second is true, by looking at mu_P at lower values)

marthaisabelhilton commented 5 years ago

Hello and thank you for your reply.

  1. I've checked and I have ~400000 events in the training sample for both MC and data. Is this ok?
  2. I actually made cuts on the data to make sure there is no regions without MC (after I made these plots)
arogozhnikov commented 5 years ago
  1. yes, that should be more than enough