Skip to content

Draft: First implementation of CatBoost in the Topo

Nicole Schulte requested to merge Ni_topological_b_cherry into master

First Draft of the topological-b classification using Catboost. For that we used the inital TMVA implementation and adjusted it to Catboost.

For now, the Topo is optimized using the HLTEfficiencyChecker to an output rate of Incl: 101.13594723341883 +/- 5.77 kHz. The Training of the model is performed on samples, that are not properly HLT1 filtered and still include around 64% of the input rate into HLT1. Therefore we are not able to provide detailed numbers on efficiencies until we are able to train on the right samples.

Th0R Implementation will come for the Neural Network Topo, that will replace the CatBoost Implementation very soon.

Done:

  • First Twobody implementation
  • First Threebody implementation
  • Build Json Reader
  • check efficiency and rates
  • validate 3body

To Do:

  • implement with ThOr

Rec!2770 (closed) lhcb-datapkg/ParamFiles!40 (closed)

CC: @mvesteri @poluekt @nnolte @msaur @gciezare @bldelane

Edited by Nicole Schulte

Merge request reports