LHCb Tracking Parametrization Training

Scripts and logics to train tracking models for the ultra-fast simulation of the LHCb experiment.

The ultra-fast simulation of the LHCb experiment is based on a combination of simple parametrization and machine-learning models that concur to the parametrization of the overall response of the whole detector.

We organize in this package the code to train and validate the machine-learning models used to parametrize the response of the tracking system.

This includes:

• the geometrical acceptance of the detector
• the tracking efficiency
• the resolution on the position and the momentum of the reconstructed tracks
• the covariance matrix of the track parameters in its closest approach to the beam direction

See also

• lb-pidgan-train for the training of the models describing the PID

Dependencies

• NumPy
• Pandas
• Uproot [scikit-hep/uproot4]
• TensorFlow 2
• Scikit-learn
• HTML Reports [html-reports]
• scikinC [scikinC]

How to train the models

The pipeline is described in a Snakefile so

snakemake -j <number-of-cores> all 

should be enough to train the whole set of models.