Choose Your Diffusion: Efficient and flexible ways to accelerate the diffusion model in fast high energy physics simulation

• DOI: 10.48550/arxiv.2401.13162
• Publication Date: 2024-01-01
• Authors: C. L. Jiang; S. J. Qian; H. Qu
• Keywords: Speedup; Benchmark (surveying)

The diffusion model has demonstrated promising results in image generation, recently becoming mainstream and representing a notable advancement for many generative modeling tasks. Prior applications of the both fast event detector simulation high energy physics have shown exceptional performance, providing viable solution to generate sufficient statistics within constrained computational budget preparation High Luminosity LHC. However, these suffer from slow generation with large sampling steps face challenges finding optimal balance between sample quality speed. study focuses on latest benchmark developments efficient ODE/SDE-based samplers, schedulers, convergence training techniques. We test public CaloChallenge JetNet datasets designs implemented existing architecture, performance generated classes surpass previous models, achieving significant speedup via various evaluation metrics.