The current neutrino detectors have been able to detect neutrinos in the range of TeV to 100 PeV, however, ultra high energy (UHE) neutrinos above 100 PeV still remain to be detected. A new neutrino detector, the RNO-G, is currently being constructed in Greenland with the purpose of detecting the first UHE neutrinos using radio antennas capable of measuring the Askaryan pulse generated after a neutrino interaction with the ice molecules. To reconstruct the neutrino's properties from the antennas' output deep learning models have been used previously. In this work we present a Graph Neural Network capable of reconstructing the shower energy and neutrino direction with similar performance compared to Convolutional Neural Networks used in previous works, using a fraction of the training data. Additionally, an increase of the reconstruction performance is shown when using the full data set.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:uu-517791 |
| Date | January 2023 |
| Creators | Serra Garet, Arnau |
| Publisher | Uppsala universitet, Högenergifysik |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | FYSAST ; FYSMAS1226 |
Page generated in 0.5329 seconds