Datasets for fitting trajectories of elementary particles using deep learning
收藏资源简介:
Training and testing datasets of simulated elementary particles, used for fitting the trajectories of elementary particles in dense materials immersed in a magnetic field using deep learning. Once decompressed, the directory structure is the following:datasets.zip*: training (1,762,327 particles in total): proton: 414,824 particles. pion: 432,855 particles. muon: 446,858 particles (muons and antimuons). electron: 467,790 particles (electrons and positrons). testing (1,759,491): proton: 412,092 particles. pion: 432,807 particles. muon: 447,003 particles (muons and antimuons). electron: 467,589 particles (electrons and positrons). checkpoints.zip: checkpoint_rnn.pth: PyTorch weights of the RNN trained model. checkpoint_transformer_encoder.pth: PyTorch weights of the RNN trained model. *This dataset corrects a bug present in previous versions related to the true initial position and momentum of the particles, which were not used in this study; therefore, the trajectory fitting results reported in the manuscript are unaffected.
本数据集为模拟基本粒子训练与测试数据集,用于采用深度学习方法拟合浸没于磁场的致密介质中基本粒子的运动轨迹。数据集压缩包datasets.zip解压后的目录结构如下: training(总计1,762,327个粒子): proton(质子):414,824个粒子 pion(π介子):432,855个粒子 muon(μ子):446,858个粒子(包含μ子与反μ子) electron(电子):467,790个粒子(包含电子与正电子) testing(总计1,759,491个粒子): proton(质子):412,092个粒子 pion(π介子):432,807个粒子 muon(μ子):447,003个粒子(包含μ子与反μ子) electron(电子):467,589个粒子(包含电子与正电子) checkpoints.zip包含以下文件: checkpoint_rnn.pth:训练完成的循环神经网络(Recurrent Neural Network, RNN)PyTorch权重文件 checkpoint_transformer_encoder.pth:训练完成的Transformer编码器PyTorch权重文件 *本数据集修复了此前版本中存在的与粒子真实初始位置及动量相关的缺陷;由于本研究未使用上述参数,因此论文中报道的轨迹拟合结果不受影响。



