[NeurIPS 2025] GraphKeeper: Graph Domain-Incremental Learning via Knowledge Disentanglement and Preservation

This repository is the official implementation of "GraphKeeper: Graph Domain-Incremental Learning via Knowledge Disentanglement and Preservation" accepted by the Main Technical Track of NeurIPS-2025.

Environment

python==3.8.0
cuda==11.7
torch==1.7.1
DGL==0.6.1

Quick Start

Unzip datas in the ROOT directory :

tar -xzvf datas.tar.gz

To train and evaluate our method, run the following command in the ROOT directory :

bash train_domain.sh

Citation

If you find this repository helpful, please consider citing the following paper. We welcome any discussions with guozh@buaa.edu.cn.

Acknowledgements

Part of this code is inspired by CGLB and TPP. We owe sincere thanks to their valuable efforts and contributions.