Heidelberg 2022 – scientific programme

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

T: Fachverband Teilchenphysik

T 79: Data Analysis, Information Technology and Artificial Intelligence 4

T 79.6: Talk

Wednesday, March 23, 2022, 17:30–17:45, T-H38

Implementing a graph-based approach for semi-inclusive tagging in Belle II — Florian Bernlochner, •Axel Heimeroth, William Sutcliffe, and Ilias Tsaklidis — Physikalisches Institut, University of Bonn, Germany

In the Belle II experiment pairs of B-mesons are produced from electron-positron collisions. The clean experimental environment allows for constraining kinematically the second B-meson if the other one is correctly reconstructed. The current tagging algorithm used in Belle II, called Full Event Interpretation, has relatively low tagging efficiency since all the intermediate decays of the second B-meson must be explicitly reconstructed. This can be critical for searches of extremely rare decays. A semi-inclusive approach, where one reconstructs only partially the tag-side, instead of an exclusive one, can be used in order to increase the overall efficiency. In this work we explore how graph neural networks can improve the purity and efficiency of a semi-inclusive approach, where only a charmed hadron, instead of a B-meson, is reconstructed. In this talk I present a proof of concept on a generic phasespace dataset and a realistic application on B −> D^* l ν decays from the official Belle II Monte Carlo.