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

T: Fachverband Teilchenphysik

T 42: Neutrino astronomy II

T 42.5: Vortrag

Dienstag, 16. März 2021, 17:00–17:15, Tq

Machine Learning-based Cascade Event Selection for IceCube — •Mirco Hünnefeld for the IceCube collaboration — TU Dortmund, Germany

IceCube is a neutrino detector located at the geographic South Pole, instrumenting a cubic kilometer of glacial ice. Neutrino interactions are detected via Cherenkov radiation of charged secondary particles. The two main detection channels consist of tracks, induced by charged current muon-neutrino interactions, and cascade events, which are almost spherical energy depositions. Although the selection and angular reconstruction of cascades is challenging, these events enhance IceCube's capabilities to probe the southern neutrino sky. In this talk, a machine learning-based cascade event selection is presented. The event selection utilizes a series of convolutional neural networks and boosted decision trees for classification and reconstruction tasks. In addition, a novel reconstruction method based on a hybrid approach of maximum-likelihood estimation and generative neural networks is employed. The presented event selection improves upon the performance of previous selections, while greatly reducing the necessary computation time, enabling the application in real-time.