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

T: Fachverband Teilchenphysik

T 42: Neutrino astronomy 2

T 42.5: Talk

Tuesday, March 5, 2024, 17:00–17:15, Geb. 30.23: 6/1

Detector Response Parametrization with Symbolic Regression for Next-Generation Neutrino Telescopes — •Arsenije Arsenić and Christian Haack — Erlangen Centre for Astroparticle Physics (ECAP), Friedrich-Alexander-Universität Erlangen-Nürnberg

Symbolic regression is a machine learning-based tool used to find mathematical expressions which describe a set of datapoints. It differs from normal regression methods in that it does not require a predefined form of the expression, but rather explores the space of various different mathematical expressions.

Neutrino telescopes are observatories for the detection of high-energy neutrinos. These telescopes are typically placed in cubic-kilometer-scale volumes of transparent material (such as ice or water) where sensors capture photons produced by neutrino interactions.

We aim to utilize symbolic regression for the geometry optimization of next-generation neutrino telescopes. As end-to-end Monte Carlo simulations are computationally expensive, symbolic regression offers an alternative by parameterizing the detector response as a function of detector layout. This helps with optimization of the detector configuration, enhances computational efficiency, and thus allows for rapid exploration of new designs.

Keywords: Machine Learning; Symbolic Regression; Neutrino Astronomy; Detectors; Optimization