Cora classification results with the following:
- Number of GCN layers: 2
- Number of hidden features: 16
- Dropout: 0.5
- Adam optimizer
- Learning rate: 1e-2
- L2 weight decay: 5e-4
- Epochs: 200
- Patience: 10
| loss | accuracy |
|---|---|
Test dataset results:
| model | loss | accuracy |
|---|---|---|
| MLP | 1.142 | 61.0 % |
| GCN | 0.961 | 71.5 % |