Skip to content

Better ways to deal with missing truth information

Dan Guest requested to merge dguest/training-dataset-dumper:fix-for-missing-truth into master

This merge request tries to unify the way we're dealing with lepton overlap removal a bit.

  • Use the same function for both the single b-tagging and double b-tagging scripts
  • Break the overlap removal into two functions, one which finds the truth leptons from W/Z, another one which does the overlap removal with jets.
  • Remove part where we call null pointers to truth particles "not a lepton". I don't like making physics decisions based on thinning. Now it throws an exception instead. In the single-btag case we catch these exceptions log them, and skip the event.
  • Also check for a null pointer in the recursive is_from_WZ function. If there's a null pointer throw an exception.

I tested this on a few thousand FTAG1 events and saw no differences in the output. @mguth or @miochoa, any comments / things to check?

Merge request reports