This repository serves as a curated collection of state-of-the-art deep graph clustering methods, encompassing papers, code, and datasets. It aims to provide researchers and practitioners with a comprehensive resource for exploring and implementing advanced techniques in graph clustering.

How It Works

The collection categorizes deep graph clustering methods into several key approaches: LLM-based, New-architecture, Temporal, Unknown Cluster Number, Reconstructive, Adversarial, and Contrastive. Each category highlights novel algorithms and their corresponding implementations, facilitating the study of diverse methodologies for uncovering underlying graph structures and node groupings.

Quick Start & Requirements

1. Download Datasets: Obtain datasets from Google Drive or specific URLs provided in the dataset tables.
2. Unzip: Extract datasets to the ./dataset/ directory.
3. Configure: Modify main.py to specify the dataset type and name.
4. Run: Execute main.py to run the clustering algorithms.

Dependencies include Python and PyTorch. Specific dataset requirements (e.g., graph vs. non-graph data) are detailed in the README.

Highlighted Details

• Comprehensive coverage of deep graph clustering techniques, categorized by methodology.
• Includes links to papers and code implementations for numerous SOTA methods.
• Provides a variety of benchmark graph and non-graph datasets, with options to construct graphs via KNN.
• Offers utility functions for data loading, preprocessing, and clustering evaluation.

Maintenance & Community

The repository is maintained by yueliu1999 and welcomes contributions. Key papers and related "Awesome" repositories are linked, suggesting an active research focus.

Licensing & Compatibility

The repository itself does not specify a license. Individual papers and code implementations will have their own licenses, which may vary. Users should verify compatibility for commercial or closed-source use.

Limitations & Caveats

This repository is a collection of external resources; it does not provide a unified framework or single executable for all methods. Users must individually set up and run each method's code. The README does not explicitly state the primary programming language or framework for all included code snippets, though PyTorch is implied by dataset loading utilities.