This is the source code for paper ''DeCaf: A Causal Decoupling Framework for OOD Generalization on Node Classification'' to appear In Proceedings of the 28th International Conference on Artificial Intelligence and Statistics (AISTATS 2025).

Xiaoxue Han, Huzefa Rangwala, Yue Ning

The code has been successfully tested in the following environment. (For older versions, you may need to modify the code)

• Python 3.8.13
• PyTorch 1.12.1+cu11.6
• pygsp 0.5.1
• sklearn 1.1.2

Please run following commands for training and testing under the src folder. We take the dataset kindle with GCN as backbone GNN model as the example.

Evaluate the TACO model

python -W ignore train.py --dataset cora --gnn GCN --ood_type label --hidden 64

Please cite our paper if you find this code useful for your research:

BibTeX

@misc{han2024decafcausaldecouplingframework,
      title={DeCaf: A Causal Decoupling Framework for OOD Generalization on Node Classification}, 
      author={Xiaoxue Han and Huzefa Rangwala and Yue Ning},
      year={2024},
      eprint={2410.20295},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2410.20295}, 
}