Göttingen 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 76: Data, AI, Computing, Electronics VII (Generative AI, MC Generators)
T 76.6: Talk
Thursday, April 3, 2025, 17:30–17:45, VG 2.101
Generative transformers for learning point-cloud simulations — Joschka Birk1, Frank Gaede2, Anna Hallin1, Gregor Kasieczka1, Martina Mozzanica1, and •Henning Rose1 — 1Institute for Experimental Physics, Universität Hamburg, Hamburg — 2Deutsches Elektron-Synchrotron DESY, Hamburg
We successfully demonstrate the use of a generative transformer for learning point-cloud simulations of electromagnetic showers in the International Large Detector (ILD) calorimeter. By reusing the architecture and workflow of the OmniJet-α model, this transformer predicts sequences of tokens that represent energy deposits within the calorimeter. This autoregressive approach enables the model to learn the sequence length of the point cloud, supporting a variable-length and realistic shower development. Furthermore, the tokenized representation allows the model to learn the shower geometry without being restricted to a fixed voxel grid.
Keywords: generative transformer; calorimeter simulation; point-cloud