Aachen 2019 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Machine-learning methods and computing in astroparticle physics

AKPIK 3.3: Vortrag

Mittwoch, 27. März 2019, 16:20–16:30, H06

Using ANNs to Find Anomalies in Waveforms Detected by IceCube — •Max Pernklau for the IceCube collaboration — Lehrstuhl für Experimentelle Physik Vb, TU Dortmund, Germany

Modern particle detectors such as the IceCube Neutrino Observatory produce large amounts of data. Almost all events detected are background events, so stringent cuts are made to ensure a usable signal-to-noise ratio. That means only phenomena which are actively searched for can be discovered, since events that cannot be properly reconstructed are removed from the analysis.

This talk deals with the viability of using artificial neural networks to search for anomalies in low-level waveform data produced by photomultiplier tubes. Unsupervised learners like Autoencoders are employed to detect unusual waveforms caused by e.g. double pulses or hardware artifacts. This approach seems promising as prior knowledge about the shape of the outlier waveforms is not required. Results might lead to a better understanding of possible detector artifacts.