This project aims to train a GAN to produce particle physics events. Currently development is being undertaken in the ipython notebook gan/notebooks/gan_skip_2.ipnb
The goal of this repository is to train a GAN to produce Drell-Yan to dimuon events in conditions that replicate the proton-proton collisions at Large Hadron Collider and the CMS detector. The input dataset uses Monte Carlo integration and simulation of the CMS detector using the pythia event generator and the Delphes detector simulator. The dataset can be pulled from here.
Once the dataset is loaded, you can launch your jupyter session and run the code in gan_skip_2.ipynb, a variety of configuration options are available which change the network architcture and training model. With the default configuration, you can expect the following performance level:
