This is a Tensorflow 2 implementation to GAN [1] LHC events.

• Use of a LorentzVector layer to implement on-shell conditions.
• Adds JS-Regularizer to the discriminator objective (Roth et al, 2017).
• Adds maximum mean discrepancy (MMD) to capture resonances.

Paper

This repository contains the code to reproduce the results shown in the paper:

• "How to GAN LHC Events" (2019) SciPost Phys. 7, 075 (2019), 1907.03764 [hep-ph].

A more detailed explaination has been given at

• ML4 Jets 2020.

Installation

Dependencies

Package	Version
Python	>= 3.7
Tensorflow	>= 2.1.0
Numpy	>= 1.15.0
wget	>= 3.2

Download + Usage

# clone the repository
git clone https://github.com/ramonpeter/EventGAN.git
# then download the datasets
cd EventGAN
python datasets/get_data.py
# Run the code
python train_gan cards/PARAM_CARD.yaml

Citation

If you use this code, please cite:

@article{Butter:2019cae,
    author = "Butter, Anja and Plehn, Tilman and Winterhalder, Ramon",
    title = "{How to GAN LHC Events}",
    eprint = "1907.03764",
    archivePrefix = "arXiv",
    primaryClass = "hep-ph",
    doi = "10.21468/SciPostPhys.7.6.075",
    journal = "SciPost Phys.",
    volume = "7",
    number = "6",
    pages = "075",
    year = "2019"
}

[1]	From ‘to GAN’, in close analogy to the verbs taylor, google, and sommerfeld.