Key points are not available for this paper at this time.
The IceCube Neutrino Observatory is located at the South Pole, which is an important tool for studying high-energy neutrinos, and provides novel insights into the universe's most energetic processes. In this study, a new reconstruction technique for IceCube events, called "dynedge," based on Graph Neural Networks (GNNs) is presented. Simulated low-energy data are used to assess the technique, with an emphasis on atmospheric neutrino oscillations in the 1 GeV–30 GeV energy range. Across the low-energy spectrum, the suggested dynedge method shows notable improvements in reconstruction accuracy for the T/C and v/μ classifications. With respect to typical atmospheric vμ–v oscillation studies, dynedge exhibits a notable 15–20% improvement in reconstructing important parameters, such as energy, zenith, direction, and interaction vertex, in the critical energy range of 1 GeV–30 GeV.
Vijayakumar et al. (Wed,) studied this question.