Graph Neural Networks for low-energy event classification & reconstruction in IceCube

Journal article

During the public Kaggle competition "IceCube -- Neutrinos in Deep Ice", thousands of reconstruction algorithms were created and submitted, aiming to estimate the direction of neutrino events recorded by the IceCube detector. Here we describe in detail the three ultimate best, award-winning solutions. The data handling, architecture, and training process of each of these machine learning models is laid out, followed up by an in-depth comparison of the performance on the kaggle datatset. We show that on cascade events in IceCube above 10 TeV, the best kaggle solution is able to achieve an angular resolution of better than 5 degrees, and for tracks correspondingly better than 0.5 degrees. These performance measures compare favourably to the current state-of-the-art in the field

Authors: Abbasi, R; Ackermann, M; Adams, J; Aggarwal, N; Aguilar, JA

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: IOP Publishing
• Journal: Journal of Instrumentation
• Volume: 17
• Issue: 11
• Pages: P11003-P11003
• Publication date: 2022-11-04
• DOI: 10.1088/1748-0221/17/11/p11003
• EISSN: 1748-0221
• ISSN: 1748-0221
• Language: English
• Keywords: Particle physics; Event (particle physics); Detector; Neutrino; Computer science; Point cloud; Optics; COSMIC cancer database; Physics; Event reconstruction; Astrophysics; Artificial intelligence