HardGAT
DGL Implementation of h/cGAO paper.
This DGL example implements the GNN model proposed in the paper HardGraphAttention.
HardGANet implementor
This example was implemented by Ericcsr during his Internship work at the AWS Shanghai AI Lab.
The graph dataset used in this example
The DGL's built-in CoraGraphDataset. Dataset summary:
- NumNodes: 2708
- NumEdges: 10556
- NumFeats: 1433
- NumClasses: 7
- NumTrainingSamples: 140
- NumValidationSamples: 500
- NumTestSamples: 1000
How to run example files
In the MVP4ModelExample folder, run
python main.pyIf want to use a GPU, run
python main.py --gpu 0If you want to use more Graph Hard Attention Modules
python main.py --num_module <your number>If you want to change the hard attention threshold k
python main.py --k <your number>Performance
TODO: Debug the implementation
TODO: Compare Cora Performance
TODO: Compare Performance in other Node classification ds
TODO: Implement Graph Classification Pipeline