Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 21: Data analysis, Information technology I

T 21.6: Vortrag

Montag, 15. März 2021, 17:15–17:30, Tu

Hadronic Shower Separation in Five Dimensions using Machine Learning Methods — •Jack Rolph, Gregor Kasieczka, and Erika Garutti — Institut für Experimentalphysik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, Deutschland

Accurate clustering of hadronic energy depositions plays a critical role in the particle flow approach proposed for future linear colliders. The highly-granular CALICE Analogue Hadronic Calorimeter prototype (AHCAL), designed with this task in mind, is distinguishable due to its ability to measure the development of a hadron shower in time as well as space. The benefit of time as an additional observable to the clustering of the simulated energy depositions of a charged and 'faked' neutral hadron observed with the AHCAL was studied using several state-of-the-art neural network architectures. These neural networks were optimised using simulations of perfect and expected operating time resolutions. As a control, networks with the same architectures were also trained without time. The clustering performance of each network relative to the control was then assessed over a range of possible operating time resolutions. For all studied networks and resolutions, the improvement in energy resolution due to time was found to be minor to negligible using these existing methods.