Simple Graph Convolutional Neural Network

---

Contents

1. Highlights
2. Requirements
3. Usage
4. Results

---

Highlights

This project is an implementation from scratch of a simple 2-layer GCN on the CORA dataset. Graph convolutional networks work by using standard convolutional layers to learn localized representations of nodes.

---

Requirements

pip install -r requirements.txt

---

Usage

To replicate the reported results, clone this repo

cd your_directory git clone git@github.com:jordandeklerk/Simple-GNN.git
cd Simple-GNN

and run the main training script

python main.py 

---

Results

We tested our simple GNN on the CORA citation dataset

• CORA
  • Simple GNN - 81.5% accuracy on the test data

Model Summary:

+----------------------+----------------------------------+----------------+----------+
| Layer                | Input Shape                      | Output Shape   | #Param   |
|----------------------+----------------------------------+----------------+----------|
| Net                  |                                  | [2708, 7]      | 23,063   |
| ├─(crd)CRD           | [2708, 1433], [2, 10556], [2708] | [2708, 16]     | 22,944   |
| │    └─(conv)GCNConv | [2708, 1433], [2, 10556]         | [2708, 16]     | 22,944   |
| ├─(cls)CLS           | [2708, 16], [2, 10556], [2708]   | [2708, 7]      | 119      |
| │    └─(conv)GCNConv | [2708, 16], [2, 10556]           | [2708, 7]      | 119      |
+----------------------+----------------------------------+----------------+----------+

---

Citations

@inproceedings{kipf2017semi,
  title={Semi-Supervised Classification with Graph Convolutional Networks},
  author={Kipf, Thomas N. and Welling, Max},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2017}
}