SANE-search-to-aggregate-neighborhood-for-gnn

Overview

This is the code for our paper Search to aggregate neighborhood for Graph Neural Networks, publised in ICDE 2021. It is a differentiable architecture search for graph neural network (GNN).

Requirements：

torch-cluster==1.5.7
torch-geometric==1.6.3
torch-scatter==2.0.6
torch==1.6.0
scikit-learn==0.21.3
numpy==1.17.2
hyperopt==0.2.5
python==3.7.4

Instructions to run the experiment

Step 1. Run the search process, given different random seeds. (The Cora dataset is used as an example)

python train_search.py  --data Cora --fix_last True  --epochs 20

The results are saved in the directory exp_res, e.g., exp_res/cora.txt.

Step 2. Fine tune the searched architectures. You need specify the arch_filename with the resulting filename from Step 1.

python fine_tune.py --data Cora  --fix_last True   --hyper_epoch 50  --arch_filename exp_res/cora.txt   

Step 2 is a coarse-graind tuning process, and the results are saved in a picklefile in the directory tuned_res.

Cite

Please kindly cite our paper if you use this code:

@inproceedings{zhao2021search,
  title={Search to aggregate neighborhood for graph neural network},
  author={Zhao, Huan and Yao, Quanming and Tu, Weiwei},
  booktitle={ICDE},
  year={2021}
}

Acknowledgement

The code is built on DARTS, one of the most well-known differentiable architecture search methods. Besides, we thank Lanning Wei, a research intern in our group, to help implement several experiments and further improve the codes in this work.

Misc

If you have any questions about this project, you can open issues, thus it can help more people who are interested in this project. We will reply to your issues as soon as possible. You are also welcomed to reach us by zhaohuan@4paradigm.com