Deep Learning for Arrival Angle Prediction in the Baikal Neutrino Telescope

This study focuses on the reconstruction of neutrino direction in the Baikal-GVD experiment using convolutional neural networks and graph neural networks. Monte Carlo simulation data are utilized, examining single-cluster events of atmospheric neutrinos with energies ranging from 10 GeV to 100 TeV. The performance of the proposed models is evaluated against the standard reconstruction algorithm by comparing their median angular resolutions. The results show that neural networks offer enhanced accuracy over the standard algorithm, particularly, in small polar angles.