Speaker
Description
The last decade has seen tremendous advances in astroparticle physics and neutrino detection. The IceCube neutrino observatory has successfully measured the cosmic neutrino flux and identified several new sources. The next generation of neutrino telescopes aim at even more energetic neutrinos in the EeV range. In-ice radio detection is a promising approach to tap into the rapidly decreasing neutrino flux at these ultra-high energies. To get the most out of these new detectors I have developed a deep learning-based reconstruction algorithm using normalizing-flows. The reconstruction can predict the neutrino energy, direction, and flavor from simulated neutrino signals. As normalizing-flows can predict the full posterior pdf for every single event, the reconstruction does not only give the best-fit values for the predicted quantities but also their uncertainties. These uncertainties can be non-Gaussian for both the neutrino direction and energy.
