HENCE-X

This is the official implement of the paper HENCE-X: Toward Heterogeneity-agnostic Multi-level Explainability for Deep Graph Networks.

Installation

Clone the repository
Create the env and install the requirements

$ git clone https://github.com/Gori-LV/HENCE-X
$ cd HENCE-X
$ source ./install.sh

Usage

Download the required datasets to /data
Download the checkpoints to /checkpoints
Run the searching scripts with corresponding dataset.

$ source ./scripts.sh

The hyper-parameters used for different datasets are shown in this script.

Examples

Run *.ipynb files in Jupyter Notebook or Jupyter Lab. We provide examples on how to use HENCE-X to explain individual instances, and show the semantic meanings of output explanation in DBLP, IMDB and MUTAG dataset, respectively.

Citation

Feel free to use our code and keep up with our progress, we kindly request you to cite our work.

@article{lv2023hencex,
  title={HENCE-X: Toward Heterogeneity-Agnostic Multi-Level Explainability for Deep Graph Networks},
  author={Lv, Ge and Zhang, Chen Jason and Chen, Lei},
  journal={Proceedings of the VLDB Endowment},
  volume={16},
  number={11},
  pages={2990--3003},
  year={2023},
  publisher={VLDB Endowment}
}