GANs for simulation of electromagnetic showers in the ATLAS calorimeter. Details of my project are in the file Report.pdf
- MNIST.py
First implementation of Improved Wasserstein GAN (WGAN-gp), on the MNIST dataset. See the article : https://arxiv.org/abs/1704.00028
- WGAN for electromagnetic showers in the ATLAS calorimeter A first script - PreProcessing.py - applies transformations to the former dataset (noise cuts, remove events>300 GeV), as describe in the file report.pdf.
Three Python files that work together :
- config.py (contains training parameters, path to files, ...)
- plot_functions.py (where all plot functions are defined)
- training.py (main file to train the WGAN)
To launch a training, you should write in the terminal : python training.py Name (Name = name of your folder in which all plots and weights will be saved. Will create the folder if it doens't already exist)
Plots are generated automatically each 250 epochs, a folder is created each time.
- A Jupyter notebook called Physics_variates.ipynb
Interactive plots