Code for Higher Order Structures in Graph Explanations

The codebase is built with Python 3.10.14.

To create a new conda environment with the required python version, use the following command:

conda create -n forge python=3.10.14

Please use the below command to install the required packages in their specific versions:

pip install -r requirements.txt

Running the code

Multiple scripts are provided to run the experiments. The scripts are located in the experiments directory. There are both real-world and synthetic datasets present in the real_world and synthetic directories respectively. The scripts are named as <dataset>_<gnn>_<explainer>.sh.

For example, to run the experiments on the benzene dataset using the gcn model and the gnn_explainer explainer, run the following command:

sh experiments/graph_classification/real_world/benzene/benzene_gcn_gnn_explainer.sh

The results will be saved in the explanations directory, under appropriate sub-directories. The results are saved in json files, with separate json files for baseline and FORGE results (graphs.json and complexes.json respectively). containing per-seed as well as average results.