Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 93: Data, AI, Computing, Electronics VIII
T 93.4: Vortrag
Freitag, 20. März 2026, 09:45–10:00, KH 00.024
Exploring Efficient Sampling Strategies for Generative Calorimeter Simulation — •Chitresh Gehloth1, Martina Mozzanica1, Thosten Buss1,2, Gregor Kasieczka1, and Frank Gaede2 — 1University of Hamburg — 2DESY, Hamburg
Detector simulation is a major computational bottleneck in high-energy physics. Generative machine-learning models, such as diffusion models and continuous normalizing flows, offer a promising alternative but require many function evaluations during sampling, leading to slow inference.
We explore trainable, non-stationary bespoke solvers that significantly reduce the number of function evaluations while preserving generation fidelity. Our study evaluates these efficient sampling strategies in the context of generative calorimeter simulation, demonstrating their potential to enable faster and more scalable detector simulations.
Keywords: calorimeter simulation; generative ml models; sampling strageties; Highly Granular Calorimeters; particle showers