Heidelberg 2022 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 107: Data Analysis, Information Technology and Artificial Intelligence 5

T 107.6: Talk

Thursday, March 24, 2022, 17:30–17:45, T-H39

PID with Recurrent Neural Networks in the ATLAS Transition Radiation Tracker for Run 3 — •Lena Herrmann, Christian Grefe, Philip Bechtle, and Klaus Desch — Physikalisches Institut, University of Bonn

The measurement of transition radiation effects by the ATLAS transition radiation tracker (TRT) is a key ingredient to the electron identification, especially at low momenta. A recurrent neural network (NN) was developed to combine hit- and track-level information into a single classifier, which significantly improves the particle identification capabilities provided by the TRT.

Since the gas configuration in the TRT will change for the upcoming Run 3 data taking period, separate RNNs have to be trained. The optimisation and training of the RNN will be presented and differences between the Run 2 and Run 3 networks will be discussed. Furthermore, the RNN response on real data taken during Run 2 will be compared to its performance in simulation.