Disentangling Homophily and Heterophily in Multimodal Graph Clustering

Brief Description

This repository contains the official source code for the paper presented at ACMMM 2025.
If you find this repository useful to your research or work, it would be greatly appreciated if you could ⭐ star this repository. ❤️

Preprint (arXiv): https://arxiv.org/pdf/2507.15253
Published Version: https://dl.acm.org/doi/pdf/10.1145/3746027.3755631

---

Requirements

The code is tested with the following environment (conda):

# Python 3.9 environment
# Core packages (conda/pip)
appdirs==1.4.4
numpy==1.26.4
pandas==2.0.3
scipy==1.13.1
scikit-learn==1.5.2
matplotlib==3.9.2
seaborn==0.13.2
torch==1.13.1+cu116
torch-geometric==2.2.0
dgl==1.0.0+cu116
faiss-gpu==1.8.0
timm==1.0.15
einops==0.8.1
kmeans-pytorch==0.3
gym==0.26.2
ogb==1.3.6
wandb==0.15.0

Citation

If you use DMGC in your work, please cite our paper (Bibtex below).

@inproceedings{guo2025disentangling,
  title={Disentangling homophily and heterophily in multimodal graph clustering},
  author={Guo, Zhaochen and Shen, Zhixiang and Xie, Xuanting and Wen, Liangjian and Kang, Zhao},
  booktitle={Proceedings of the 33rd ACM International Conference on Multimedia},
  pages={2044--2053},
  year={2025}
}

Datasets

DMGC Data Repository