#  We plan to release the code with GPLv3 license, but currently the code might contain some legacy code that has no license. So we do not include a license, but reviewers are free to view and run the project to confirm the reproducibility.
#  The code is very messy. Please do not distribute. We will release cleaned-up project after publishing the paper.
#  CFVAE was the code name of our project :)
#  cfvae_art.py, cfvae_ihdp.py, cfvae_pokec.py contain entry code for the three experiment sections, respectively.
#  IHDP dataset is not included due to its size, and it can be obtained from https://www.fredjo.com.
#  The preprocessed Pokec dataset is saved as adj.pickle.
#  The python environment can be recovered from environment.yml, using conda. The main dependencies are TensorFlow and Edward, both are quite *old* versions.
#  [For 'Towards...' paper] We updated the Code for a parallel submission and also fixed several minor bugs after making the figures and tables in experiments, so the results might be slightly different if you run the current Code.

