Qualification Task for Maxime Fernoux

Maxime Fernoux (CPPM, Marseille) will be responsible for the development of the Flavour Tagging tagger DIPS in release 21.9 and study its performance in the context of the ITk upgrade simulation for the HL-LHC. The goal here is to set up and test the training procedures on release 21.9 samples. An optimization of the sets of inputs to DIPS will be performed with a particular focus on the tagger performance in the |eta| > 2.5 region, which has not been considered in Run 2 optimization efforts. Particular attention will be given to the hyperparameter optimisation of the DeepSet network considering the lower statistics available in Upgrade samples in comparison to Run 2-3 MC samples. Maxime will contribute to the maintenance and the development of the framework to train the DIPS taggers.
Maxime will report his progress in the Flavour Tagging software and algorithms meetings and the corresponding software will be made available in GitLab and adequately documented in the ATLAS Software Documentation.

Start date: 1 October 2021
Local supervisor: duperrin
Additional local support: tstreble
Technical supervisor: biliu

CC: vdao, mguth, pgadow, duperrin, thuffman