sycny / ENGAGE

Code for our ECML'23 paper 'ENGAGE: Explanation Guided Data Augmentation for Graph Representation Learning'

ENGAGE

This repository hosts the code for our ECML'23 paper 'ENGAGE: Explanation Guided Data Augmentation for Graph Representation Learning', by Yucheng Shi, Kaixiong Zhou, Ninghao Liu.

Appendix

ENGAGE_Appendix.

Dependencies

• torch 1.10.1+cu113
• torch-cluster 1.5.9
• torch-geometric 2.0.3
• torch-scatter 2.0.9
• torch-sparse 0.6.12
• faiss-cpu 1.7.2

If you have trouble in installing torch-geometric, you may find help in its official website.

Training & Evaluation

Graph-level

For SimCLR model:

python Simclr_graph_self_guided.py --dataset DD

For Simsiam model:

python Simsiam_graph_self_guided.py --dataset DD

Node-level

For SimCLR model:

python Simclr_self_guided.py --dataset Cora --model GAT

For Simsiam model:

python Simsiam_self_guided.py --dataset Cora --model GCN

Acknowledgements

Parts of implementation are reference to GRACE, GraphCL, and Simsiam.