README.md

Precise simulation of electromagnetic calorimeter showers using a Wasserstein Generative Adversarial Network

Code for the training of a generative model that generates electron-induced, electromagnetic showers in the CMS HGCAL prototype that was tested in October 2017 at the CERN H2 beam line.

References

• Computing and Software for Big Science: https://link.springer.com/article/10.1007%2Fs41781-018-0019-7
• preprint on arXiv: https://arxiv.org/abs/1807.01954

Recommended hardware prerequisites

• NVIDIDA GPU GTX 1080 or better
• 4GB RAM (for loading the training data)

Software prerequisites

The code has been tested with:

• Python 2.7
• numpy v 1.16 or newer
• ROOT 5 or 6
• Tensorflow v1.5 with keras and GPU support (recommended)

Input data files

Input data files for the training are provided on the CERN computing infrastructure:

• /afs/cern.ch/work/t/tquast/public/Sept2017_HGCALTB_Sim

Running the code

• Training command: python training.py --EpochStart 0 --Nepochs 150 --checkpoint_dir <directory where network files are stored> --input_dir <Directory where the input files are> (training duration for 150 epochs with recommended hardware ~30h)