Calibrating the ParticleNet tagger for 2023 CMS data

Machine learning algorithms have provided the CMS collaboration with exciting new ways to identify the sources of jets. The ParticleNet algorithm is a particular powerful multiclassifier that can identify jets from top quarks and W bosons with high accuracy, as well as jets from many other sources. This talk will present the calibration process for top quark and W boson identification criteria using ParticleNet. We will share how the selection criteria are determined using multijet simulation, as well as the technique for fitting a top quark or W boson mass peak in events that either pass or fail a specific selection criterion. These fits enable CMS to determine correction factors so that the selection efficiency in simulation matches that observed in data, which is critical to enabling usage of this algorithm by the collaboration.