Erlangen 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 50: Data, AI, Computing, Electronics V

T 50.6: Talk

Wednesday, March 18, 2026, 17:30–17:45, KH 00.024

Normalizing-Flow-Based Reweighting of Detector Systematics in Neutrino Telescopes — •Oliver Janik — Erlangen Centre for Astroparticle Physics (ECAP), Friedrich-Alexander-Universität Erlangen-Nürnberg

Accurate measurements of the astrophysical neutrino flux require reliable predictions of the detector response in neutrino telescopes. Because this response is highly non-analytical, such predictions rely on Monte Carlo (MC) simulations and a forward-folding approach. A key limitation is that MC simulations must assume specific detector properties. Evaluating detector systematic uncertainties therefore requires either simulating multiple MC sets with fixed detector parameters or using approaches such as SnowStorm, which sample detector parameters continuously on an event-by-event basis. In both cases, reweighting typically relies on interpolation between MC expectations, introducing an implicit dependence on an assumed flux model.We present a normalizing-flow*based approach that factorizes detector systematics into shape and yield components. A conditional normalizing flow is used to model changes in the distribution of reconstructed observables, while the overall event yield is modeled separately as a function of the relevant detector parameters. This separation enables consistent reweighting for both discrete MC sets and SnowStorm simulations, without relying on flux-dependent interpolation.In this talk, we demonstrate the application of this method to detector-systematics modeling in astrophysical neutrino flux measurements.

Keywords: IceCube; KM3NeT; Detector Response; Normalizing Flows