Uploaded image for project: 'ATLAS Flavour Tagging'
  1. ATLAS Flavour Tagging
  2. AFT-634

Improve Umami Preprocesing



    • Task
    • Resolution: Unresolved
    • Minor
    • None
    • None
    • None


      The student (ioleksiy) will work on improving the current pre-processing chain for the machine-learning based flavour tagging algorithms. This work will be carried out within the FTAG Algorithm group. It consists of the following aspects:

      • he will start fixing pre-processing-related issues in the umami framework implementing full unit and integration tests for the preprocessing part. The preprocessing needs to be rewritten in a more modular way and needs to be prepared such that the preprocessing base module can be pulled out of umami
      • The preprocessing should be adapted such that beam spot weights are taken into account.
      • Neutral PFlow object will be implemented in the Athena workflow, such that they are available in the training-dataset-dumper. Whenever possible the code shall be written such that it is portable for large-R jet. Afterwards, he will evaluate how much the neutral particles information helps the tagger training. For this, the latest version of the b-tagging algorithm at that time should be used.

      The complete workflow will be maintained and documented in git, making sure all the different pre-processing methods are documented.

      The work will be reported in the Algo meeting regularly and be documented in git (SW) and as an internal note (performance studies).

      532770 Performance Studies - Flavour Tagging
      556629 Performance Studies for Flavour Tagging

      Local supervisor: golling
      Thecnical supervisor: fdibello




            ioleksiy Ivan Oleksiyuk
            fdibello Francesco Armando Di Bello
            0 Vote for this issue
            6 Start watching this issue