M. Milesi ; M. Hohmann ; P. Urquijo ; A. Narimani Charan ; A. Novosel ; L. Santelj ; P. Križan ; T. Ferber

20 October 2022

Abstract: We present a new model for lepton identification with the Belle II electromagnetic calorimeter (ECL). In the central, barrel region, we use energy-weighted CsI(Tl) crystal images in a convolutional neural network architecture to learn energy deposition patterns of electrons, muons and charged hadrons. In the forward and backward angular regions of the ECL, we exploit higher-level observables associated to the lateral energy spread, and per-crystal pulse shape discrimination information in a set of boosted decision trees trained categorically.

Keyword(s): Particle identification ; ECL ; BDT ; CNN ; Calorimeter clustering