Refining MC-based Jet Energy Scale Calibrations with Machine Learning at the ATLAS Experiment

Hadronic jets are important for many measurements and searches for new physics at the LHC. Current procedures for calibrating the energy of hadronic jets for the ATLAS Experiment require multiple, time-intensive stages and are restricted in the number of parameters that can be considered. This talk presents results on a mixture density neural network-based approach using Monte Carlo simulations of QCD processes, allowing studies into an expanded parameter space and consideration of correlations between parameters. The machine learning approach is expected to be able to combine several steps in the calibration and to improve the robustness of the calibration to pile-up from additional proton-proton collisions.