neutrino-man4/case_2023

Variational Autoencoders for the CMS Anomaly Search Effort (CASE)(MLG-23-002 and EXO-22-026):

Don't bother trying to understand how all of it works. It just does.

Be sure to install Python3, Tensorflow 2.X, h5py, scikit-learn and the usual packages. If something's missing, you will know anyways.

Getting Started

The paths are stored as key-value pairs in the form of Python dictionaries present in the following scripts (only necessary ones to get it all running are mentioned below):

• pofah/path_constants/sample_dict.py

  • base directory where all samples are stored: base_dir_events

• pofah/path_constants/sample_dict_file_parts_input.py

  • same as above: base_dir
  • directories where individual samples are stored: sample_dir
    • qcdSigTrain: Data on which the VAE is to be trained (likewise for test data)
    • qcdSigMCOrig: The original MC signals before sampling is carried out to replace it with QCD jets having similar kinematics.

• pofah/path_constants/sample_dict_file_parts_reco.py

  • writing the reconstructed files to disk: base_dir (these files are moderately large so watch your disk quota)

• pofah/path_constants/experiment_dict.py

  • saving trained model: model_dir
  • rest should be self-explanatory.

Once you are done setting the paths, move to vande/. Don't look at the other folders for now.