Hamburg 2016 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 74: Experimentelle Methoden II
T 74.3: Talk
Wednesday, March 2, 2016, 17:15–17:30, VMP8 SR 105
Identification of Hadronic Tau Decays at the ATLAS Detector Using Artificial Neural Networks — Dirk Duschinger, Stefanie Hanisch, Wolfgang Mader, •Nico Madysa, and Arno Straessner — Institut für Kern- und Teilchenphysik, TU Dresden, Germany
One of the primary goals of the ATLAS experiment at the LHC is the search for physics beyond the Standard Model. The efficient identification of hadronically decaying tau leptons is crucial for this as they comprise the final states of several decay channels sensitive to new physics. (e. g. Higgs boson decays H → τhad τhad) The identification algorithm currently applied at ATLAS utilizes multi-variate methods and reconstructed particle properties to discriminate against QCD jets, which constitute an important background.
This talk presents a new neural-network-based approach to hadronic tau decay identification and investigates its dependence on hyperparameters such as the network topology or number of training cycles. Ensembling is presented as a technique to improve classifier performance and robustness against overtraining. The resulting classifier is compared to the current approach based on Boosted Decision Trees. The study is based on 2012 data taken at the ATLAS detector at a center-of-mass energy of √s = 8 TeV.