Hunting for vampires and other unlikely forms of parity violation at the Large Hadron Collider: calo-image datasets for the standard model

An example calo-image dataset used in the paper: standard model.

Each shard is named `calo-image_sm_${DATASET}_${INDEX}.tar.gz`. Each contains one data file. DATASET is in {train,test,private_test} to label the three independent splits for training, validation, and testing respectively. INDEX labels separate batches which should be trivially combined.

Each data file is in h5 format. Its data are under the key "entries" as an array with shape (n, 32, 32) corresponding to (event_index, eta, phi) for the n calorimeter images in an unrolled eta--phi surface.

本论文所使用的示例量能器图像（calo-image）数据集，基于标准模型（standard model）构建。

每个数据分片的命名格式为`calo-image_sm_${DATASET}_${INDEX}.tar.gz`，每个分片仅包含一个数据文件。其中${DATASET}的取值为{train, test, private_test}，分别对应训练、验证与测试三个独立数据集划分；${INDEX}用于标记独立批次，各批次可直接合并使用。

所有数据文件均采用HDF5（h5）格式存储，数据以键"entries"存储为形状为(n, 32, 32)的数组：其中n为该文件包含的量能器图像总数，数组各维度依次对应事件索引（event_index）、赝快度（eta）与方位角（phi），对应展开的η-φ平面上的n幅量能器图像。