Optimization of the neural network trigger for a detection of cosmic rays in surface detectors of the pierre auger observatory

One of the greatest challenges for nowadays astrophysics is to understand the origin of the ultrahigh-energy cosmic rays (UHECR). Possibility of detection of air showers initiated by neutrinos can significantly help to find sources of the UHECR. Detection technique, however, requires very sophisticated algorithm due to very low cross section of neutrinos. Our algorithm is based on a shape recognition by artificial neural networks (ANN). It can efficiently separate air showers initiated very deep in the atmosphere (“young” showers - which can be potentially induced by neutrinos) from air showers which started at the edge of the atmosphere (“old” showers). The algorithm uses a significant amount of resources, so it has been implemented in the largest Cyclone
^®
V E FPGA with many Digital Signal Processing blocks. MATLAB tests shows that size of the ANN can be decreased, which saves not negligible amount of FPGA resources.