OmniJet-α_C: learning point cloud calorimeter simulations using generative transformers

Abstract

Abstract\n We show the first use of generative transformers for\n generating calorimeter showers as point clouds in a high-granularity\n calorimeter. Using the tokenizer and generative part of the\n OmniJet-α model, we represent the hits in the\n detector as sequences of integers. This model allows variable-length\n sequences, which means that it supports realistic shower development\n and does not need to be conditioned on the number of hits. Since the\n tokenization represents the showers as point clouds, the model\n learns the geometry of the showers without being restricted to any\n particular voxel grid.

Joschka Birk et al., OmniJet-α_C: learning point cloud calorimeter simulations using generative transformers, J. Inst.