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

T: Fachverband Teilchenphysik

T 16: Experimental Methods (general) 1

T 16.4: Talk

Monday, March 21, 2022, 17:00–17:15, T-H29

Jet Vertex Tagger in release 22 — •Abdullah Nayaz¹, Teng Jian Khoo², and Cigdem Issever³ — ¹Humboldt University, Berlin, Germany — ²Humboldt University, Berlin, Germany — ³Humboldt University, Berlin, Germany

Pile-up mitigation is a crucial part of many important Particle Physics analysis e.g. HH->4b. The Jet Vertex Tagger (JVT) is a multivariate pile-up suppression variable developed for the ATLAS experiment that combines information from other track based pile-up variables and plays a major role in ATLAS analysis. In this study, as part of the preparation for Run 3 data-taking and analysis, the performance of JVT has been checked for the new release 22 Track to Vertex Association (TTVA) working points using Monte-Carlo simulated dijet data samples. First, the TTVAs that result in a good performance of the JVT have been identified. Furthermore, to increase the JVT performance, a Multilayer Perceptron Neural Network (NN) has been used to retrain the JVT for release 22. The training was done separately for offline and trigger level jets, varying the inputs to the NN to optimise the separation of hard scatter and pile-up jets. Some improvement on the JVT performance was observed after the training process which will be beneficial for Run 3 ATLAS analyses.