Münster 2017 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 23: Experimentelle Techniken der Astroteilchenphysik 2
T 23.5: Vortrag
Montag, 27. März 2017, 17:45–18:00, S 055
Online Classification of IceCube Events using Neural Networks — •Joshua Luckey for the IceCube collaboration — Technische Universität Dortmund, Deutschland
The IceCube neutrino detector is located at the geographic South Pole and consists of 5160 digital optical modules, each containing a photomultiplier tube, deployed into the ice. With an instrumented volume of 1 km2 IceCube detects events at a rate of about 3000 Hz. The first data processing steps are done by a system of online filters, which are applying reconstruction algorithms to the data. An analysis on the data at this early stage bears the advantage of being independent of time- and CPU-intensive data preprocessing. In this talk a classification of online data of the IceCube detector is presented. The classification is based on the topology of the events in the detector. At first the events can be separated into the two classes of track-like and cascade-like events and from there further classifications can be carried out. With the recent advancements in other fields of research in mind, Deep Learning algorithms in conjunction with neural networks are used to conduct the afore-mentioned classification as early in the data acquisition process as possible. A classification at this early stage could be beneficial to analyses focusing on just one type of event. Furthermore an optimization of the used neural net, with the aim of minimizing the classification time, could be performed to classify every detected event.