Speaker
Description
The calibration of the energy scale and resolution of jets, the collimated sprays of particles initiated by quarks and gluons, is important for many precision measurements and searches for physics beyond the standard model at the Large Hadron Collider (LHC). Currently within ATLAS, a series of calibrations is required to correct jets for effects of pileup and detector response. This results in several (often large) corrections with a loss of correlations between the steps and artificial constraints. ATLAS is exploring new approaches for jet calibration based on machine learning (ML) that can, in principle, perform many of the corrections in one step and address the limitations of the current approach. This is particularly relevant for developing jet calibrations for physics performance studies at the future High Luminosity LHC, where there will be 3-4 times more pileup from additional pp collisions. In this talk, a ML-based approach to jet calibration will be presented using simulated samples of the upgraded ATLAS detector at the HL-LHC. Data formatting procedures, network structure/modifications, metrics for performance, and future extensions will be discussed.
