dmlc--dgl
Chang Liu
6e1be69a84
* Refactor gat example * Add ppi support * Minor update * Update * Update * Change valid_xxx to val_xxx * Readme Update * Update Co-authored-by: Mufei Li <mufeili1996@gmail.com>
923 B
923 B
Graph Attention Networks (GAT)
- Paper link: https://arxiv.org/abs/1710.10903
- Author's code repo (tensorflow implementation): https://github.com/PetarV-/GAT.
- Popular pytorch implementation: https://github.com/Diego999/pyGAT.
How to run
Run with the following for multiclass node classification (available datasets: "cora", "citeseer", "pubmed")
python3 train.py --dataset cora
Run with the following for multilabel classification with PPI dataset
python3 train_ppi.py
NOTE: Users may occasionally run into low accuracy issue (e.g., test accuracy < 0.8) due to overfitting. This can be resolved by adding Early Stopping or reducing maximum number of training epochs.
Summary
- cora: ~0.821
- citeseer: ~0.710
- pubmed: ~0.780
- ppi: ~0.9744