Skip to content

Radiative inclusive hh(h)g lines to ThOr

Aniol Lobo Salvia requested to merge alobo_b_radinclusive into master

Lines

  • Hlt2RD_BToHHGamma_Incl_Line - Included (running in ThOr only)

not yet included in all_lines (still needing 1 loki functor, running now in ThOr but with one dummy input for the BDT)

  • Hlt2RD_BToHHHGamma_Incl_Line
  • Hlt2RD_BToHHGamma_GammaToEE_Incl_Line
  • Hlt2RD_BToHHHGamma_GammaToEE_Incl_Line

Missing ThOr functors:

  • MINTREE/MAXTREE (used in BDT inputs in hhhg[min IP, max TRCHI2], hhgee and hhhgee [min PT])
  • TODO urgent:
    • Add a missing cut to hh combiner -> Depends on Rec!2810 (merged) and seems pretty irrelevant since it's not filtering any candidates: cut on IP of K+ forming hh was already applied to K+ selection
    • Remove TrackChi2, GhostProb cuts, DD KS and Lambda
    • use RD prefilters
    • Tune rate
    • Decide on persistance
    • Check that vertex variables are ok -> solved in Rec!2811 (merged) , B vertex CHI2DOF < 1000 is a safety cut at 100: probably not affected either way
  • Optimization for a future MR
    • Rethink builders: use more RD builders, 2-3 hadron combinations from Topo builders
      (waiting for them to be ported to ThOr)
    • Retrain BDTs with added Ks0 and L0 and correct Truth Matching. Pending:
      • Find a way to substitute the input needing MINTREE
      • Apply truth matching to neutrals on training samples. (was giving problems in the past)
      • Process the wide list of MC samples used for training at preselection level. (it takes a while)
      • Retrain on real run 3 data when available
    • Isolation variables: apply them when going back to persisting only extra particles for a better selection.

Since the lines are already working ok and optimizations may take a while to be ready, it's has been considered better to push the line as it is for now and leave the rest for future improvements.

Rates and efficiencies

Checked on full upgrade_minbias_hlt1_filtered (roughly 15 000 events) for different values of the BDT cut (nominal is 0.18)
Statistics is not very good, e.g. 43 passed events for nominal cut, 5 of them exclusive KstG efficiency tested on MC evtNumber 11102202 ,prodID 75829

Line Incl Excl KstG eff
Hlt2RD_BToHHGamma_Incl_LineDecision Incl: 2.869 +/- 0.43 kHz Excl: 0.33 kHz 0.119 +/- 0.011
Hlt2RD_BToHHGamma_bdt_cut02_Incl_LineDecision Incl: 2.336 +/- 0.39 kHz, Excl: 0.33 kHz -
Hlt2RD_BToHHGamma_bdt_cut022_Incl_LineDecision Incl: 1.935 +/- 0.35 kHz Excl: 0.33 kHz 0.113 +/- 0.011
Hlt2RD_BToHHGamma_bdt_cut024_Incl_LineDecision Incl: 1.601 +/- 0.32 kHz Excl: 0.33 kHz 0.111 +/- 0.011
Hlt2RD_BToHHGamma_bdt_cut025_Incl_LineDecision Incl: 1.468 +/- 0.31 kHz Excl: 0.27 kHz 0.106 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut026_Incl_LineDecision Incl: 1.134 +/- 0.27 kHz Excl: 0.20 kHz 0.106 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut028_Incl_LineDecision Incl: 0.867 +/- 0.24 kHz Excl: 0.20 kHz 0.101 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut03_Incl_LineDecision Incl: 0.6007 +/- 0.20 kHz Excl: - kHz 0.099 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut032_Incl_LineDecision Incl: 0.400 +/- 0.16 kHz Excl: 0.07 kHz 0.094 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut034_Incl_LineDecision Incl: 0.266 +/- 0.13 kHz Excl: 0 kHz 0.089 +/- 0.01
Hlt2RD_BToHHGamma_bdt_cut036_Incl_LineDecision Incl: 0.133 +/- 0.09 kHz Excl: 0 kHz 0.081 +/- 0.009
Hlt2RD_BToHHGamma_bdt_cut038_Incl_LineDecision Incl: 0.133 +/- 0.09 kHz Excl: 0 kHz 0.073 +/- 0.009
Hlt2RD_BToHHGamma_bdt_cut04_Incl_LineDecision Incl: 0.0667 +/- 0.06 kHz Excl: 0 kHz -

logs:

K1G_HHHGamma_Moore.log KstG_HHGamma_Moore.log KstIsoG_HHGamma_Moore.log Lambda_HHGamma_Moore.log MinBias_HHGamma_Moore.log MinBias_HHHGamma_Moore.log

Persistance

From last dicussion: we will run as an HLT2 line persisting the reconstruction - meaning if I understood correctly that we will need to go in the sprucing stream. Motivation is to be able to tune cone variables for the extra particles. Extra selected particles (multiplicities in !1038 (merged) ) still included since it would be useful to test the framework, can be deleted while persistreco=True if need be.

Edited by Aniol Lobo Salvia

Merge request reports