Introduction

One of my colleagues asked me: "What's your most practical skill in network science?" I told him:" community detection"! ！

Whether you are a network science beginner, enthusiast, or expert, whether you study network data or other data networking, community detection technology will be able to accompany you at every stage of network analysis --- from Network Data Preprocessing and Analyse, Network Visualization, Network Advanced Insight Acquisition...

Community detection can be an organic part of your own models! The VGAER we developed provides an opportunity to combine cutting-edge GNN methods.

Come and try!

VGAER

Simple and efficient -- a novel unsupervised community detection with the fusion of modularity and network structure:

Get to Start--VGAER

Overview

This repository contains the implementation of the VGAER (Variational Graph Auto-Encoder with Reinforcement learning) model. The model is designed to learn node representations in a graph and perform community detection. It uses a combination of graph neural networks and variational autoencoders to achieve this.

Requirements

To run the code, You can install these packages using pip:

pip install -r requirements.txt

Installation

To use the VGAER model, clone the repository and navigate to the directory:

git clone https://github.com/your-repo/vgaer-model.git
cd vgaer-model

Usage

Command-Line Arguments

The model can be configured using the following command-line arguments:

• --model: The type of model to use (default: gcn_vae)
• --seed: Random seed for reproducibility (default: 42)
• --epochs: Number of epochs to train (default: 10000)
• --hidden1: Number of units in the first hidden layer (default: 8)
• --hidden2: Number of units in the second hidden layer (default: 2)
• --lr: Initial learning rate (default: 0.05)
• --dropout: Dropout rate (default: 0.0)
• --dataset: Type of dataset to use (default: cora)
• --cluster: Number of communities to detect (default: 7)

Running the Model

To run the model, use the following command:

python main.py --model gcn_vae --dataset cora --cluster 7

This will train the VGAER model on the Cora dataset with 7 communities.

Code Structure

• model.py: Contains the implementation of the VGAER model.
• cluster.py: Contains community detection algorithms.
• NMI.py: Contains the Normalized Mutual Information (NMI) score calculation.
• Qvalue.py: Contains the Q value calculation for community detection.
• main.py: The main script that loads the dataset, trains the model, and performs community detection.

License

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

Contributing

Contributions to the project are welcome. Please open an issue or submit a pull request with your proposed changes.

Acknowledgements

This project was inspired by various research papers and open-source projects in the field of graph neural networks and community detection. We gratefully acknowledge their contributions to the field.

We also thank the willleeney/ugle project for referencing and integrating our VGAER model. We look forward to engaging with more open-source projects to jointly advance the development and application of technology.

Requirement

dgl==0.8.0.post1

matplotlib==3.5.1

networkx==2.7.1

numpy==1.22.3

pandas==1.4.1

scikit_learn==1.0.2

scipy==1.8.0

seaborn==0.11.2

torch==1.11.0

Citation

Please cite our paper if you use this code or our model in your own work:

@inproceedings{qiu2022VGAER,
title={VGAER: Graph Neural Network Reconstruction based Community Detection},
author={Qiu, Chenyang and Huang, Zhaoci and Xu, Wenzhe and Li, Huijia},
booktitle={AAAI: DLG-AAAI'22},
year={2022}
}