This repository is for a simple implementation of Graph Convolutional Networks. The original paper of GCN is here.

In this repository, there are 2 datasets provided to train the GCN model:

- Citeseer
- Cora

The two datasets can be downloaded in here. The downloaded datasets should be unziped and restored into a directory named datasets. The following tree shows the path to each files in the datasets:

datasets
├── citeseer
    ├── citeseer.cites
    ├── citeseer.content
    └── README
└── cora
    ├── cora.cites
    ├── cora.content
    └── README

1. Install Python 3.

2. Install the Python packages in requirements.txt. If you are using a virtual environment for Python package management, you can install all python packages needed by using the following bash command:

   $ pip install -r requirements.txt

3. Install PyTorch. The version of PyTorch should be greater or equal than 1.7.0. This repository provides the CUDA usage.

1. Modify config.json as your machine setting. The following explanations are for understanding train_config of config.json:

   • num_epochs: The number of epochs of the training process. Default: 200
   • learning_rate: The learning of the optimizer for the training process. Default: 0.01

2. Execute training process by train.py. An example of usage for train.py are following:

   $ python train.py --dataset_name=cora

   The following bash command will help you:

   $ python train.py -h

• Initially complemented the implementations.

• Refactoring.
• Write docstrings.

• Semi-supervised Classification with Graph Convolutional Networks
• Revisiting Semi-Supervised Learning with Graph Embeddings