Graph-Convolutional-Networks

This is a PyTorch implementation of graph convolutional networks (GCNs) from the paper Semi-Supervised Classification with Graph Convolutional Networks.

Usage

In the src directory, edit the config.json file to specify arguments and flags. Then run python main.py.

Limitations

• Does not support mini-batch training.
• Currently, only supports the Cora dataset. However, for a new dataset it should be fairly straightforward to write a Dataset class similar to datasets.Cora.

References

• Semi-Supervised Classification with Graph Convolutional Networks, Kipf and Welling, ICLR 2017.
• Collective Classification in Network Data, Sen et al., AI Magazine 2008.