Speaker
Description
The Belle II experiment aims to search for physics beyond the Standard Model by producing a large number of B meson pairs with the SuperKEKB accelerator and precisely measuring rare decays and CP violation. The TOP (Time-Of-Propagation) counter is the particle identification detector in the barrel region of Belle II, which distinguishes kaons from pions using the time of arrival and position of Cherenkov photons. In this study, we develop an identification method based on a graph neural network (GNN), which takes the detected Cherenkov photons as input, aiming to improve the K/π identification performance. Currently, we are constructing and training the GNN model using Belle II Monte Carlo simulation data, and comparing its identification performance with that of the conventional method. In this poster, we present an overview of the GNN-based identification method and the current results of its performance evaluation, and discuss future prospects, including the application of more advanced models such as the Particle Transformer and the application to real data.