Towards Faster Deep Graph Clustering via Efficient Graph Auto-encoder

An official code for paper "Towards Faster Deep Graph Clustering via Efficient Graph Auto-Encoder". This repository is based on the unified framework for deep attribute graph clustering (https://github.com/Marigoldwu/A-Unified-Framework-for-Deep-Attribute-Graph-Clustering) .

Requirements

The proposed FastDGC is implemented with python 3.10.12 on a Tesla T4 GPU.

Python package information is summarized in requirements.txt:

matplotlib==3.5.3
munkres==1.1.4
numpy==1.21.5
scikit_learn==1.0.2
torch==2.3.0

Quick Start

Pre-training for new dataset

python main.py -P -M pretrain_egae_for_FastDGC -D cora -LS 1 -S 325

• The pre-training code file is located at ./model/pretrain_egae_for_FastDGC/train.py.
• Pre-trained datasets [acm, dblp, cora, cite, amap, wisc, texas], which are saved at ./pretrain/pretrain_egae/FastDGC/.

Fine-tuning for pre-trained dataset

python main.py -M FastDGC -D acm -LS 10

• You can modify the data set through the -D command. Optional data sets are [acm, dblp]. If you want to run other datasets, please go to the website to download.
• You can use -TS to obtain the t-SNE visualization of clustering results, and the picture will be saved to the 'img/clustering/FastDGC'. Note that LS should be set to 1, otherwise a picture will be drawn for each experiment
• You can use -H to obtain the heatmap of embeddings, and the picture will be saved to the 'img/heatmap/FastDGC'. Note that LS should be set to 1, otherwise a picture will be drawn for each experiment

Colab Notebook

Or you can download and unzip the code to Google Drive, name it FastDGC, and then run it in Colab. https://colab.research.google.com/drive/1u89965L3AejZVWVlE1B70MfqNESdtvtj?usp=sharing

Our paper is just accepted by ACM Transactions on Knowledge Discovery from Data.

Citation

If you use our code, please cite our paper as:

@article{ding2024towards,
author = {Ding, Shifei and Wu, Benyu and Ding, Ling and Xu, Xiao and Guo, Lili and Liao, Hongmei and Wu, Xindong},
title = {Towards Faster Deep Graph Clustering via Efficient Graph Auto-Encoder},
year = {2024},
issue_date = {September 2024},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
volume = {18},
number = {8},
issn = {1556-4681},
url = {https://doi.org/10.1145/3674983},
doi = {10.1145/3674983},
journal = {ACM Trans. Knowl. Discov. Data},
month = {aug},
articleno = {202},
numpages = {23},
keywords = {Deep graph clustering, graph auto-encoder, graph neural networks, unsupervised learning}
}