hgcal_wgan

WGANs for HGCAL data at CMS.

Project developed during my summer working at CERN's Openlab with Dr. Maurizio Pierini.

Most of the work is documented in my report (although it misses some updates).

Dependencies:

• NumPy
• matplotlib
• h5py
• setGPU
• gpustat
• tensorflow-gpu
• keras (v. >= 1.2.0)
• root_numpy

Structure

• env contains scripts to set up the environment at SF's GPUs.
• preprocessing contains tools to preprocess ROOT files and convert them to HDF5.
• gans contains GAN implementations for training. The best so far is wgan_conv2d_regression.py.
• notebooks contains Jupyter Notebooks used for evaluation.