Göttingen 2025 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Machine Learning in Particle- and Astroparticle Physics

AKPIK 3.5: Vortrag

Donnerstag, 3. April 2025, 17:15–17:30, Theo 0.134

Adaptive Generative Modeling for Accelerated Calorimeter Simulations via Domain Transfer — •Lorenzo Valente¹, Fank Gaede², Gregor Kasieczka^1,3, and Anatolii Korol² — ¹Institut für Experimentalphysik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany — ²Deutsches ElektronenSynchrotron DESY, Notkestr. 85, 22607 Hamburg, Germany — ³Center for Data and Computing in Natural Sciences CDCS, Deutsches Elektronen-Synchrotron DESY, Notkestr. 85, 22607 Hamburg, Germany

Simulating particle collider detectors presents significant computational challenges, with current methods struggling to scale with increasingly complex experimental datasets. Deep generative models offer a promising solution for dramatically reducing computational overhead, especially as upcoming particle physics experiments are expected to produce unprecedented volumes of data.

We introduce a novel domain adaptation framework that utilises state-of-the-art deep generative models to generate high-fidelity 3D point-cloud representations of particle showers. Using transfer learning techniques, our approach adapts simulations across diverse electromagnetic calorimeter geometries with exceptional data efficiency, thereby reducing training requirements and eliminating the need for a fixed-grid structure.

Preliminary results demonstrate that our method can achieve high accuracy while significantly reducing data and computational demands, offering a scalable solution for next-generation particle physics simulations.

Keywords: Deep Generative Models; Calorimeter Simulations; Transfer Learning; Diffusion Models; Normalising Flow