gat

Graph Attention Networks (GAT)

• Paper link: https://arxiv.org/abs/1710.10903
• Author's code repo (in Tensorflow): https://github.com/PetarV-/GAT.
• Popular pytorch implementation: https://github.com/Diego999/pyGAT.

Dependencies

• tensorflow 2.1.0+
• requests

pip install tensorflow requests
DGLBACKEND=tensorflow

How to run

Run with following:

python3 train.py --dataset=cora --gpu=0

python3 train.py --dataset=citeseer --gpu=0 --early-stop

python3 train.py --dataset=pubmed --gpu=0 --num-out-heads=8 --weight-decay=0.001 --early-stop

Results

Dataset	Test Accuracy	Baseline (paper)
Cora	84.2	83.0(+-0.7)
Citeseer	70.9	72.5(+-0.7)
Pubmed	78.5	79.0(+-0.3)

• All the accuracy numbers are obtained after 200 epochs.
• All time is measured on EC2 p3.2xlarge instance w/ V100 GPU.