DGCLUSTER: A Neural Framework for Attributed Graph Clustering via Modularity Maximization

This repository is a reference implementation of DGCluster as described in the paper:

DGCLUSTER: A Neural Framework for Attributed Graph Clustering via Modularity Maximization.
Aritra Bhowmick, Mert Kosan, Zexi Huang, Ambuj Singh, Sourav Medya
Association for the Advancement of Artificial Intelligence (AAAI), 2024.
https://ojs.aaai.org/index.php/AAAI/article/view/28983

Requirements

• Install Torch from https://pytorch.org/get-started/locally/
• Run install.py to install the required packages after installing Torch.

python install.py

or

• Install the required packages from env.yml file with conda.

conda env create -f environment.yml

Reproducing Results

To generate the result for a specific dataset (e.g., cora) and lambda parameter (e.g., 0.2), run the following:

python main.py --dataset cora --lam 0.2

We also provide shell scripts run.sh, run_alp.sh, run_base_all.sh to reproduce all results data for all datasets and parameters. You can just run the following commands to generate the data for the corresponding tables or figures.

• ./run.sh: Table 2, Table 3, and Figure 1.
• ./run_alp.sh: Table 5.
• ./run_base_all.sh: Table 4.

Additionally, plots.py and plots_num_clusters.py can be used to generate Figure 1 and Figure 2.