IDGC: Interpretable Deep Graph-level Clustering

This repository contains the implementation of IDGC (Interpretable Deep Graph-level Clustering), a novel approach for interpretable and efficient graph clustering. Our work has been accepted for presentation at ICPR 2024 (International Conference on Pattern Recognition).

Overview

IDGC is a deep learning-based framework designed for clustering graph-structured data. It leverages graph neural networks (GNNs) combined with interpretable clustering techniques to provide meaningful insights into complex graph data.

Requirements

Before running the code, ensure you have installed the necessary dependencies:

conda create -n idgc_env python=3.8
conda activate idgc_env

# Install PyTorch and PyTorch Geometric
pip install torch==1.8.0 torch-geometric==2.0.2

# Additional dependencies
conda install -c pytorch faiss-cpu=1.7.4 mkl=2021 blas=1.0=mkl
pip install matplotlib
pip install pandas
pip install soyclustering
pip install rdkit

Usage

To run IDGC, execute the following command:

python -m models.train

Configuration

You can modify the configuration settings in utils/Configures.py according to your dataset and experimental setup.

Citation

If you find this work useful in your research, please consider citing:

@inproceedings{your_paper_reference,
  title={Interpretable Deep Graph Clustering},
  author={Jianbin Cui, Lingyang Chu},
  booktitle={Proceedings of the 2024 International Conference on Pattern Recognition (ICPR)},
  year={2024}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgements

We would like to thank the community for their valuable contributions and support.