jabarr (DESY/UCL) will contribute to various tasks in FTAG. On the algorithm side, he will maintain the UMAMI and training dataset dumper (TDD) packages during his QT. This task includes adding new features when needed, expanding the use of automated software testing/validation and providing up-to-date documentation. The SW will be maintained in a git repo and the documentation can be either detailed README or a twiki.
In addition, he will explore different WP definitions to achieve better performance at high pT. This should in principle include: Constructing fixed b-efficiency WPs and flat light-rejection WPs, as functions of jet pT, and check/compare the performance in various samples (Note: this is well defined and has been done before). Exploring alternative definitions besides the fixed efficiency/rejection rate strategy (Note: straightforward but outcome is unknown, considered as R&D).
On the calibration side, he will work on:
Contribute to the Run 2 r22 di-lepton ttbar calibration for the newest generation of taggers (DL1d as the primary goal while GNN as an optional choice) in preparation for undertaking the first Run 3 data calibration. Using early Run 3 data, Jackson will firstly validate the modelling of the b-tagging and related variables using e-mu + 2-jet ttbar dominated data samples (similar to the studies conducted with the early Run 2 data documented in ATL-PHYS-PUB-2015-039), and if sufficient Run 3 data is collected will also undertake a preliminary Run 3 calibration.