Karlsruhe 2024 – scientific programme

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

T: Fachverband Teilchenphysik

T 126: Top physics 5 (top mass)

T 126.4: Talk

Friday, March 8, 2024, 09:45–10:00, Geb. 30.95: Audimax

Mass-Decorrelated Classification of Unlabeled Data for tt Identification — •Sofia Brozzo, Patrick Connor, Johannes Lange, Peter Schleper, and Hartmut Stadie — Institut für Experimentalphysik, Universität Hamburg

Precision measurements of the top quark mass are an important tool to test the Standard Model. Although the fully hadronic decay channel provides the largest branching fraction, the large QCD multijet background leads to a challenging event selection. To improve on the mass resolution and reduce combinatorical tt and QCD background, a kinematic fit is applied before the top quark mass is extracted.

Here, a neural network trained on unlabeled CMS data is employed to further improve the selection of tt events and to reject QCD background.

To further ensure that the neural network is not biased on the top mass, the aim of this analysis is to decorrelate the neural network output from the input mass via distance correlation.

Keywords: Machine Learning; Top; CWoLa; Distance Correlation