Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

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

T: Fachverband Teilchenphysik

T 91: Machine Learning: Event and jet reconstruction

T 91.3: Vortrag

Freitag, 3. April 2020, 11:30–11:45, H-HS I

Study on the use of convolutional neural networks for strange-tagging based on jet images from calorimeters — •Nils J. Abicht, Johannes Erdmann, Olaf Nackenhorst, and Sonja Zeissner — TU Dortmund, Lehrstuhl für Experimentelle Physik IV

In addition to already existing algorithms for bottom- and charm-tagging, a technique that identifies jets originating from the hadronisation of strange quarks (strange-tagging) would be useful for various analyses at the LHC. This study focuses on making use of calorimeter information for this identification in the form of jet images, i.e. a representation of energy depositions in η and φ. Such jet images are built from simulations of jets from strange and down quarks. Convolutional neural networks (CNNs), which are especially geared towards extracting possible patterns in images, are used to learn the distinctive features of the strange and down jet images. During the optimization of the performance of the final CNN, different preprocessing steps as well as CNN layouts are explored in order to create a new method for strange-tagging.