brain-gcn-2

Few different projects in one repo.

• Melanie code in the brain_project directory;
• An experiment with clustering and VAEs in the VAdE directory;
• Current graph inference experiments in brain_project/nri directory.

Running NRI training

There are two different training loops:

1. Training on the real data (train.py). Parameters are set within the file directly. Data needs to be generated first (using the data_utils.py file) and placed where the trainer can see it (see the data_folder parameter).
2. Training on synthetic data (train_synthetic.py). Parameters are also set within the file, but this should run without much hassle.

Dependencies

• PyTorch (version 0.4)
• tensorboardX (pip install tensorboardX)
• The usual standard python packages numpy, matplotlib, scipy, etc.