Aachen 2019 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 77: Deep Learning III

T 77.3: Vortrag

Donnerstag, 28. März 2019, 16:30–16:45, H06

A Neural Network Approach to Estimate the Mass of Resonances decaying to τ⁺τ⁻ with the ATLAS Detector — •Martin Werres, Philip Bechtle, Klaus Desch, Christian Grefe, Michael Hübner, Lara Schildgen, and Peter Wagner — Physikalisches Institut, Uni Bonn, Deutschland

This study investigates the predictive power of deep neural networks (DNN) in the task of mass reconstruction in ditau events with simulated ATLAS data at the LHC. The ditau mass has a large discriminating power in distinguishing between Z→ττ and H→ττ events. A strategy how to design a DNN training environment is presented. The performance is compared to the Missing Mass Calculator tool [arXiv:1012.4686] which is currently used in analyses. The conditions under which a competitive mass reconstruction can be achieved are presented. The influence of the environment of the individual event in the training of the DNN, such as the tau decay mode, the rest of the event, the pileup conditions and other influences on the reconstruction are studied.