Valid, Sparse, and Stable Explanations in Graph Neural Networks

by Thorben Funke, Megha Khosla, Mandeep Rathee, and Avishek Anand

1. Requirements

See requirements.txt for the main python packages used to run this repository with python 3.7.

2. Data

The real-world datasets will be downloaded via pytorch-geometric. For the synthetic dataset, we included the file generate_gnnexplainer_dataset.py and in data/syn2.npz our resulting graph.

3. Execute Zorro

You can simply run

python3 execution.py

to get explanations for the default setting: 10 nodes for Cora and GCN with tau=0.85.

We included the save points of the model and the randomly selected nodes in the results directory.

4. Evaluation

Running evaluate_zorro.py will create csv files with the evaluate explanations.