000002323 001__ 2323
000002323 005__ 20210401115349.0
000002323 037__ $$aBELLE2-TALK-CONF-2021-015
000002323 041__ $$aeng
000002323 088__ $$aBELLE2-TALK-DRAFT-2021-028
000002323 100__ $$aKatharina Dort
000002323 245__ $$aIdentification of exotic highly ionising particles at the Belle II pixel detector using unsupervised autoencoders
000002323 260__ $$aDPG Spring Meeting 2021$$c2021-03-09
000002323 300__ $$amult. p
000002323 520__ $$aThe Belle II experiment at the high luminosity SuperKEKB e+e- collider has started operation in 2018. The present setup features a 1-layer DEPFET pixel detector with 4 M pixels and trigger rates up to 5 kHz, installed in close proximity to the interaction region. It offers the unique opportunity of detecting highly ionising particles such as antideuterons, pions with small transverse momentum < 100 MeV or exotic particle species like magnetic monopoles with a very short track length. Multivariate techniques are attractive tools to cope with the identification of these particles against the high beam background rates. In this contribution, we evaluate the performance of unsupervised autoencoders, in order to discriminate signal from beam background. In particular, we investigate the possibility of performing the training directly on background data, an unbiased approach to (a) avoid theoretical assumptions about the nature of signal and (b) avoid supervised training with Monte-Carlo data.
000002323 700__ $$aStephanie Käs
000002323 700__ $$aJens Sören Lange
000002323 8560_ $$fsoerenlange@yahoo.com
000002323 8564_ $$uhttps://docs.belle2.org/record/2323/files/BELLE2-TALK-CONF-2021-015.pdf