Fast low energy reconstruction using Convolutional Neural Networks
(2026) Journal of Instrumentation — Vol. 21 (2026)
Files
document.pdf
Open Access - Adobe PDF
- 2.12 MB
Details
- Authors
- et. al.
- Abstract
- IceCube is a Cherenkov detector instrumenting over a cubickilometer of glacial ice deep under the surface of the SouthPole. The DeepCore sub-detector lowers the detection energythreshold to a few GeV, enabling the precise measurements ofneutrino oscillation parameters with atmospheric neutrinos. Thereconstruction of neutrino interactions inside the detector isessential in studying neutrino oscillations. It is particularlychallenging to reconstruct sub-100 GeV events with the IceCubedetectors due to the relatively sparse detection units and detectionmedium. Convolutional neural networks (CNNs) are broadly used inphysics experiments for both classification and regressionpurposes. This paper discusses the CNNs developed and employed forthe latest IceCube-DeepCore oscillationmeasurements [1]. These CNNsestimate various properties of the detected neutrinos, such as theirenergy, direction of arrival, interaction vertex position,flavor-related signature, and are also used for backgroundclassification.
- Affiliations
Citations
de Wasseige, G., Genton, E., Kruiswijk, K., Lamoureux, M., Lazar, J., Raab, C., Myhr, P., Vereecken, M., IceCube, & et al. (2026). Fast low energy reconstruction using Convolutional Neural Networks. Journal of Instrumentation, 21. https://doi.org/10.1088/1748-0221/21/02/P02020 (Original work published 2026)