Skip to content

Inspection of jet-flavor tagging in full simulation at CLD.

License

Notifications You must be signed in to change notification settings

saracreates/FullSimTagger

Repository files navigation

Jet-Flavor tagging on full simulation

For the first time at FCC-ee, we perform jet-flavor tagging in full simulation. This repository hosts the code for these first studies on full simulation at CLD. Therefore, we

  1. Extract the desired variables (such as kinematics, PID, and track parameters) from CLD full simulation
  2. Compare these full simulation variables to fast simulation variables at CLD
  3. Train the Particle Transformer on this data (using weaver-core, not shown here). We also perform studies on fast simulation of CLD and IDEA.
  4. Evaluate the performance. (Check out this note and this code)

The work is published on the CERN Document Server and used for the FCC feasibility study.

Part of this code is implemented in key4hep, checkout this repo.

Overview of this repository

  • the src folders contain the code to extract all desired variables from full simulation. There are different variations such as

    • fixing the pandora algorithm by using tracks and not PFOs to describe charged particles (src_tc-match)
    • using tracks only and neglecting neutral particles (src_tracks).
    • src_trackPFOcheck does a study on lost charged particles is PFOs.
    • src_debug_tracks performs a study on the 5 helix track parameters and its covariance parameters. Each of these folders has create_jet_based_tree.py and tree_tools.py where the core of the work lies.
  • We use condor to submit the jobs for creating desired root files. The submit files (analysis.sub) for the different cases are in respective submit folders and automatically created by the write_analysis.py files. The analysis.sub files execute the submitAnalysisJob.sh script.

Example usage

  • clone this repository
  • source key4hep via source /cvmfs/sw.hsf.org/key4hep/setup.sh -r 2024-04-12
  • go into the desired src folder and run

python create_jet_based_tree.py /eos/experiment/fcc/prod/fcc/ee/test_spring2024/240gev/Hbb/CLD_o2_v05/rec/00016562/000/Hbb_rec_16562_111.root test_output.root

Citation

If you find this code helpful and use it in your research, please cite:

@manual{aumiller_2024_4pcr6-r0d06,
  title        = {Jet Flavor Tagging Performance at FCC-ee},
  author       = {Aumiller, Sara and
                  Garcia, Dolores and
                  Selvaggi, Michele},
  month        = nov,
  year         = 2024,
  doi          = {10.17181/4pcr6-r0d06},
  url          = {https://doi.org/10.17181/4pcr6-r0d06}
}

About

Inspection of jet-flavor tagging in full simulation at CLD.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published