GraphHash

This is the offical Github repository of GraphHash.

GraphHash: Graph Clustering Enables Parameter Efficiency in Recommender Systems

Xinyi Wu, Donald Loveland, Runjin Chen, Yozen Liu, Xin Chen, Leonardo Neves, Ali Jadbabaie, Clark Mingxuan Ju, Neil Shah, Tong Zhao

The ACM Web Conference (WWW) 2025

---

Install packages

conda create -n graphhash python=3.9
conda activate graphhash
pip install -r requirements.txt

---

Download data

The processed data can be downloaded from https://drive.google.com/file/d/1vcyDljpxG3-TJdzgbhVdoWIK-eA_wIF-/view?usp=sharing

Make sure the unzipped folder dataset is under the same directory as the train.py file in this repo.

Detailed notebooks containing preprocessing steps are included.

---

Run experiments

Example command to run GraphHash for the retrieval task on Gowalla with full batch size:

python train.py model=MF hash_type=graph dataset=Gowalla dataset.lr=1e-2 dataset.wd=1e-8 dataset.loss=BPR

All the commands for running our experiments are provided in commands.sh, with the best searched hyperparameters. Note that the experiments are configured using hydra. Please see its documentation for detailed usage.

---

Citation

Thank you for your interest in our work! Please consider citing

@inproceedings{Wu2025GraphHashGC,
  title={GraphHash: Graph Clustering Enables Parameter Efficiency in Recommender Systems},
  author={Xinyi Wu and Donald Loveland and Runjin Chen and Yozen Liu and Xin Chen and Leonardo Neves and Ali Jadbabaie and Clark Mingxuan Ju and Neil Shah and Tong Zhao},
  booktitle={Proceedings of the ACM Web Conference 2025},
  year={2025}
}

Please email xinyiwu@mit.edu if you run into any problems.