Aachen 2019 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 4: Deep Learning I

T 4.2: Vortrag

Montag, 25. März 2019, 16:15–16:30, H06

Further development of the ATLAS Deep Learning flavour tagging algorithm — •Manuel Guth — Albert-Ludwigs Universität, Freiburg, DE

The development of machine learning techniques is making a lot of progress in the last few years. Already now, machine learning is deeply embedded in our daily life. Especially deep neural networks require a large amount of statistics for a robust training procedure in order to find yet unknown dependencies in data. The large amount of simulated data available in particle physics allows to use these new sophisticated techniques to improve the physics analyses. The identification of heavy flavour jets (tagging) plays an important role in almost all physics analyses at the ATLAS experiment. It is an essential tool for precision measurements as well as for searches for new physics phenomena. One of the frameworks within ATLAS for b-tagging is the Deep Learning tagger (DL1). It uses deep neural networks based on TensorFlow and Keras to distinguish b-, c- and light flavour jets using the information of several baseline b-taggers. A first introduction of the DL1 tagger is given, followed by detailed studies to improve the deep learning network architecture.