Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 103: Search for Dark Matter IV

T 103.3: Vortrag

Freitag, 20. März 2026, 09:30–09:45, AM 00.014

Fast Template-Based Inference via Conditional Normalizing Flows for XENONnT — •Johannes Merz for the XENON collaboration — Institut für Physik & Exzellenzcluster PRISMA⁺⁺, Johannes Gutenberg-Universität Mainz

Template-based likelihood analysis is a cornerstone of the inference in the XENONnT experiment. It’s limited in computational efficiency and flexibility by its reliance on histogram based templates. In this talk, we present a fast inference approach that replaces traditional templates with continuous, differentiable models based on conditional normalizing flows. The models provide an accurate representation of detector response distributions while enabling efficient likelihood evaluation. This significantly accelerates and simplifies parameter inference across the parameter spaces. Our results demonstrate that normalizing flow based templates offer a scalable and efficient alternative to classical template methods for XENONnT.

Keywords: direct dark matter detection; dual-phase XENON TPC; Machine Learning Methods; Normalizing Flows