Published June 9, 2021 | Version v1
Dataset Open

Electromagnetic Calorimeter Shower Images of CaloFlow

  • 1. Rutgers University

Description

These are the calorimeter showers that were used to train and evaluate the normalizing flows of "CaloFlow: Fast and Accurate Generation of Calorimeter Showers with Normalizing Flows" and "CaloFlow II: Even Faster and Still Accurate Generation of Calorimeter Showers with Normalizing Flows". The training and evaluation scripts can be found in this git repository.

The samples were created with the same GEANT4 configuration file as the original CaloGAN samples. Said configuration can be found at the CaloGAN repository; the original CaloGAN samples are available at this DOI.

Samples for each particle (eplus, gamma, piplus) are stored in a separate .tar.gz file. Each tarball contains the following files:

  • train_particle.hdf5: 70,000 events used to train CaloFlow I and II.

  • test_particle.hdf5: 30,000 events used for model selection of CaloFlow I and II.

  • train_cls_particle.hdf5: 60,000 events used to train the evaluation classifier.

  • val_cls_particle.hdf5: 20,000 events used for model selection and calibration of the evaluation classifier.

  • test_cls_particle.hdf5: 20,000 events used for the evaluation run of the evaluation classifier.

Each .hdf5 file has the same structure as the original CaloGAN data.

Files

Files (1.1 GB)

Name Size Download all
md5:80993770abb4cddd356e708505c3c4bc
426.1 MB Download
md5:1ef4bd6d8ffe76042a998c5795da8fa3
412.8 MB Download
md5:42b61d9802723a85dee2f4fc42ec8ead
302.3 MB Download

Additional details

Related works

Is supplement to
Preprint: https://arxiv.org/abs/2106.05285 (URL)
Preprint: https://arxiv.org/abs/2110.11377 (URL)
Journal article: 10.1103/PhysRevD.107.113003 (DOI)
Journal article: 10.1103/PhysRevD.107.113004 (DOI)
Dataset: 10.5281/zenodo.7849133 (DOI)