Skip to content

HF Gluon split label

Nikita Ivvan Pondrequested to merge
npond/training-dataset-dumper:late_shower_label into main

Description

Introduces an alg which assigns a jets an additional label which is 1 if they contain a HF hadron which itself comes from gluon splitting. Gluon splitting is determined by:

  • find all b-hadrons ghost matched to a jet, if none get all c-hadrons, if none -> 0
  • Take the leading hadron, and iterate up its decay chain till we find a same flavour and charge quark
  • Iterate up the quark chain (internal mathsy stuff) until we get to a point where we don't have a same flavour and charge quark
  • If the parent is not a gluon ->
  • If the parent is a gluon, check if it has another child such that (pid1 + pid2) = 0. If yes -> return 1, else return 0

This can hopefully be useful for several reasons:

  • For training, we currently treat g->bb jets as signal. In most cases, we probably don't want this. This new label will allow us to split out b- and c- labels into b-HS, b-fromG, c-HS, c-fromG (maybe HS is the incorrect name as these would all come from the HS, but that can be handled later in upp)
  • Some analyses have large backgrounds made up of gluon splitting into -b @amaiza messaged me recently on the best way to label jets as coming from gluon splitting, and I mentioned I had this MR in the works.
  • These g->bb jets often look quite different to our normal b-jets, see plots on my previous very similar MR

Review checklist:

  • CI Passing
  • Comments addressed
  • Source branch is up to date with target

Merge request reports