The First Application of Neural Networks to the Analysis of the TUS Orbital Detector Data

The results of application of simple neural networks for recognizing events caused by charged particles hitting the photodetector of the TUS first orbital telescope of extreme-energy cosmic rays are presented. Perceptrons with different numbers of hidden layers and convolutional neural networks were employed for the task. The details of the study and encountered problems together with their solutions are provided. The results demonstrate the efficiency and good prospects of using machine learning methods for analyzing and classifying data of similar experiments. The work will be continued and extended with the TUS detector data and with the data of the Mini-EUSO experiment that is currently carried out at the Russian Segment of the International Space Station.