Abstract
We employ neural networks for classification of data of the TUS fluorescence telescope, the world’s first orbital detector of ultra-high energy cosmic rays. We focus on two particular types of signals in the TUS data: track-like flashes produced by cosmic ray hits of the photodetector and flashes that originated from distant lightnings. We demonstrate that even simple neural networks combined with certain conventional methods of data analysis can be highly effective in tasks of classification of data of fluorescence telescopes.
Highlights
Instant track-like flashes (TLFs) caused by charged particles hitting the UV filters of the photodetector; flashes produced by light coming outside of the field of view (FOV) of the detector and scattered on its mirror; they were called “slow flashes” because of the long signal rise time in comparison with TLFs; and events with complex spatio-temporal dynamics; these included so-called ELVEs, which are short-lived optical events that manifest at the lower edge of the ionosphere as bright rings expanding at the speed of light up to a maximum radius of ∼300 km [34], events with waveforms that could be expected from fluorescence originating from extensive air showers produced by extreme energy cosmic rays, as well as violent flashes of a yet unknown origin
We presented results of using two types of neural networks—multilayer perceptrons and convolutional networks—for classification of data obtained with the TUS orbital fluorescence telescope
Both multilayer perceptrons (MLPs) and convolutional neural networks (CNNs) demonstrated a high accuracy in classifying two types of events registered by TUS: instant track-like flashes produced by cosmic ray hits of the focal surface of the detector and so-called slow flashes that used to originate from distant lightnings
Summary
In the early 1980s, Benson and Linsley suggested to increase dramatically the exposure of UHECR experiments by a totally different approach They suggested to put a wide-field-of view fluorescence telescope into a low-Earth orbit and employ it for registering fluorescence and Cherenkov emission of extensive air showers born by UHECRs in the nocturnal atmosphere [6,7]. This suggests developing methods suitable for an analysis of their data, including methods based on neural networks. All neural networks were implemented in Python with the Keras library [30] available in TensorFlow [31]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
