Göttingen 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 54: Data, AI, Computing, Electronics V (Anomaly Detection, Event Selection)
T 54.4: Talk
Wednesday, April 2, 2025, 17:00–17:15, VG 2.101
Using End-to-End Optimized Summary Statistics in IceCube — •Oliver Janik and Christian Haack — Erlangen Centre for Astroparticle Physics (ECAP), Friedrich-Alexander-Universität Erlangen-Nürnberg
The characterization of the astrophysical neutrino flux with the IceCube Neutrino Observatory traditionally relies on a binned forward-folding likelihood approach. However, this method is constrained by the need for sufficient Monte Carlo (MC) statistics in each bin, which limits both the granularity and dimensionality of the binning scheme. By employing a neural network to learn a one-dimensional summary statistic, it becomes possible to optimize the binning scheme for the analysis while maintaining adequate MC statistics per bin. This, for example, allows the use of a larger number of observables in order to improve the analysis performance. The talk will go into detail on the application of end-to-end optimized summary statistics in the context of analyzing and characterizing the galactic neutrino flux.
Keywords: Binning; Optimization; IceCube