DeepMuon: Generating Cosmic-Ray Muons Based on Optimal Transport
DeepMuon: Generating Cosmic-Ray Muons Based on Optimal Transport
Wang, Mr. Ao-Bo 1Su, Mr. Bohua Pan, Mr. Chu-Cheng Dong, Dr. Xiang 2Sun, Mr. Yu-Chang Hu, Mr. Yu-Xuan Cheng, Mr. Ao-Yan 2Cai, Prof. Hao Fan, Prof. Xi-Long
作者信息
- 1. Wuhan University School of Physics and Technology
- 2. Wuhan University
- 折叠
摘要
Deep learning can learn distributions directly from data, enabling fast event simulation without hand-crafted designs or costly sampling. Yet many mainstream generators, such as diffusion models, require sophisticated architectures and multi-step denoising, which reduces throughput. Neural generators also struggle with sharp, high-kurtosis distributions common in physics, such as the cosmic-ray muon energy spectrum. Cosmic-ray muon simulation underpins non-destructive imaging, positioning and navigation, timing, and cryptography, and is traditionally performed with Monte Carlo or parametric models. Here we propose DeepMuon, a deep learning-based cosmic-ray muon generator that targets both efficiency and accuracy while addressing high-kurtosis fitting. We first apply an inverse Box--Cox transformation to reduce energy-spectrum kurtosis and simplify the learning task. We then optimize the generator with the Sliced Wasserstein Distance loss, achieving high-fidelity, one-step cosmic-ray muon generation with only a 2-layer Transformer encoder. DeepMuon also learns muon distribution patterns from limited data, enabling simulation of real detector-measured distributions. At sea level, DeepMuon substantially accelerates muon generation relative to CRY. We further develop a DeepMuon-based pipeline to directly simulate underwater muon distributions and compare against CRY, demonstrating suitability for simulation and imaging tasks. For more details about our open-source project, please visit:https://github.com/wangab0/deepmuon.
Abstract
Deep learning can learn distributions directly from data, enabling fast event simulation without hand-crafted designs or costly sampling. Yet many mainstream generators, such as diffusion models, require sophisticated architectures and multi-step denoising, which reduces throughput. Neural generators also struggle with sharp, high-kurtosis distributions common in physics, such as the cosmic-ray muon energy spectrum. Cosmic-ray muon simulation underpins non-destructive imaging, positioning and navigation, timing, and cryptography, and is traditionally performed with Monte Carlo or parametric models. Here we propose DeepMuon, a deep learning-based cosmic-ray muon generator that targets both efficiency and accuracy while addressing high-kurtosis fitting. We first apply an inverse Box--Cox transformation to reduce energy-spectrum kurtosis and simplify the learning task. We then optimize the generator with the Sliced Wasserstein Distance loss, achieving high-fidelity, one-step cosmic-ray muon generation with only a 2-layer Transformer encoder. DeepMuon also learns muon distribution patterns from limited data, enabling simulation of real detector-measured distributions. At sea level, DeepMuon substantially accelerates muon generation relative to CRY. We further develop a DeepMuon-based pipeline to directly simulate underwater muon distributions and compare against CRY, demonstrating suitability for simulation and imaging tasks. For more details about our open-source project, please visit:https://github.com/wangab0/deepmuon.关键词
deep learning/cosmic muon generator/muon tomography/muon radiography引用本文复制引用
Wang, Mr. Ao-Bo,Su, Mr. Bohua,Pan, Mr. Chu-Cheng,Dong, Dr. Xiang,Sun, Mr. Yu-Chang,Hu, Mr. Yu-Xuan,Cheng, Mr. Ao-Yan,Cai, Prof. Hao,Fan, Prof. Xi-Long.DeepMuon: Generating Cosmic-Ray Muons Based on Optimal Transport[EB/OL].(2026-02-15)[2026-02-27].https://chinaxiv.org/abs/202602.00199.学科分类
天文学/测绘学
评论