Convolutional Neural Networks for Upsampled Cosmic-Ray Radio Waveforms at the IceCube Neutrino Observatory

Extensive Air Showers occur when Cosmic Rays collide with particles in the Earth's atmosphere. These showers emit electromagnetic radiation detectable by radio antennas. A prototype station equipped with three antennas has been collecting background and air-shower data at the IceCube Neutrino Observatory at the South Pole. By combining this background with simulated cosmic-ray signals generated by the CoREAS software, we can analyze the expected waveforms specific to the South Pole environment. Traditional methods have been employed to classify and denoise radio signals from cosmic-ray events. However, Convolutional Neural Networks have outperformed these methods in the classification and denoising tasks. This study investigates the impact of upsampled waveforms on the accuracy of such Convolutional Neural Network classifiers and denoisers.