CERN Accelerating science

ATLAS Note
Report number ATL-PHYS-PUB-2024-015
Title Transformer networks for constituent-based b-jet calibration with the ATLAS detector
Corporate Author(s) The ATLAS collaboration
Collaboration ATLAS Collaboration
Publication 2024
Imprint 28 Jul 2024
Number of pages 29
Note All figures including auxiliary figures are available at https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PUBNOTES/ATL-PHYS-PUB-2024-015
In: BOOST 2024 - 16th International Workshop on Boosted Object Phenomenology, Reconstruction, Measurements, and Searches at Colliders, Genova, It, 29 Jul - 2 Sep 2024
Subject category Particle Physics - Experiment
Accelerator/Facility, Experiment CERN LHC ; ATLAS
Free keywords machine learning ; neural networks ; transformers ; energy calibration ; mass calibration ; b-jets ; JetEtMiss ; JETETMISS
Abstract The precise measurement of a jet’s kinematics is a critical component of the physics program based on proton–proton collision data recorded by the ATLAS detector at the Large Hadron Collider. The determination of the energy and mass of jets containing bottom quarks b-jets is particularly difficult as, for example, they have different radiation patterns compared to the average jet and can contain heavy-flavour decays into a charged lepton and an unobserved neutrino. This document reports on a novel calibration technique for jets focusing on b-jets using transformer-based neural networks trained on simulation samples to correct reconstructed jet properties to the true values. Separate simulation-based regression methods have been developed to estimate the transverse momentum of small-radius jets and the transverse momentum and mass of large-radius jets. In both cases, the regression methods move the median measurement closer to the true value. A relative resolution improvement with respect to the nominal calibration between 18% and 31%, depending on the transverse momentum, is demonstrated for small-radius jets. Both the large-radius jet transverse momentum and mass resolution are shown to improve by 25–35%.
Scientific contact person Cerutti, F, (fabio.cerutti@cern.ch)

Corresponding record in: Inspire


 Record created 2024-07-28, last modified 2024-07-28


  • Send to ScienceWise.info