This repository contains the necessary scripts to convert specifically formatted Root files to binary hdf5 files for training, as well as Jupyter notebooks to train a basic VAE architecture and analyse its performance with the relevant datasets. Please view the environment.yml file to download the relevant python libraries to load all the project contents (including tensorflow).

To prepare the initial data, one should use the convert_to_h5.py script. The syntax necessary to run it is as follows:

''' python convert_to_h5.py --input-file INPUT ROOT FILE NAME -- output-file PICK YOUR FAVOURITE OUTPUT FILE NAME.hdf5 '''

Note that this script expects that the root file analysed has a tree name with the relevant region data called 'l1UpgradeEmuTree/L1UpgradeTree'. If this directory is not within the ROOT file and does not contain the TTrees 'vRegionEt','vRegionEta','vRegionPhi', then it will need to be modified.

The notebook ZeroBias_VAE.ipynb contains all the steps necessary to initialise, train and test a basic VAE on pre-prepared training/testing hdf5 datasets.

We have also included another plotting script VAE_Plotting_Script.ipynb to plot the resulting training and testing data (in particular, the loss values) from the original network.