Skip to content

Topological Trigger new Tuning

Nicole Schulte requested to merge Topo_Neural_Net into master

With this tuning to the Topo, we have updated the following:

  • Introduced the corrected mass as a feature in both MVAs
  • tighten the cuts in the ParticleFilter and ParticleCombiners which should increase the speed and allow for a better fine tuning of the MVA Cut
  • Retrained the model with samples that have the same cuts applied
  • Lowered the rate

The rate was determined in the following way:

  1. The HLTEfficiencyChecker was used on the sample upgrade_minbias_hlt1_filtered from the TestFileDB
  2. A scan was performed, that loops over multiple 2body/3body combinations accounting for overlap (If an event gets selected by both lines, it is only considered once when calculating the rates)
  3. MVA Cuts were chosen based on the upgrade_minbias_hlt1_filtered sample (nominal conditions, VELO closed)
  4. Selection was applied on the new 2023 minbias corresponding to VELO open conditions to understand what rates can be expected with recent data

The rates are the following:

upgrade_minbias_hlt1_filtered

Line Rate
2body 20.04 +/- 1.5 kHz
3body 15.64 +/- 1.3 kHz
Combined 28.30 +/- 1.75 kHz

UpgradeHLT1FilteredWithGEC

Line Rate
2body 18.65 +/- 0.21 kHz
3body 12.41 +/- 0.17 kHz
Combined 23.28 +/- 0.24 kHz

2023 Minbias Velo Open

Line Rate
2body 2.0 +/- 0.82 kHz
3body 1.33 +/- 0.67 kHz
Combined 2.19 +/- 0.34 kHz

Efficiency plots will be delivered by @bldelane soon

upgrade_minbias_hlt1_filtered rates for different MVA cuts (if adjustment is desired)

2body Cut 3body Cut. Combined Rate.
0.990 0.980 38.98 +/- 2.0473 kHz
0.985 0.985 39.86 +/- 2.0698 kHz
0.985 0.980 45.15 +/- 2.1991 kHz
0.980 0.990 44.60 +/- 2.1860 kHz
0.980 0.980 51.31 +/- 2.3400 kHz

Corresponding Merge Requests: Paramfiles: lhcb-datapkg/ParamFiles!63 (merged)

CC: @bldelane @gciezare @poluekt @mvesteri

Edited by Nicole Schulte

Merge request reports