Deformable Graph Convolutional Networks

This repository is the implementation of Deformable Graph Convolutional Networks (Deformable GCNs).

Jinyoung Park, Sungdong Yoo, Jihwan Park, Hyunwoo J. Kim, Deformable Graph Convolutional Networks, In AAAI Conference on Artificial Intelligence (AAAI) 2022.

Environmental Setup

We provide the datasets via this link.

# Python version : 3.8.13, Cuda version : 10.2
$ conda env create --file env.yaml
$ conda activate deformablegcn

Running the code

Arg	Description
—dataset	Dataset
—lr	Learning rate
—weight_decay	weight decay
—epochs	Number of epochs to train
—hidden	Dimensionality of hidden embeddings
—dropout	Dropout probability
—num_blocks	Number of blocks
—n_neighbor	Number of neighbors of latent neighborhood graphs
—n_hops	Number of hops (l)
—n_kernels	Number of kernels (k)
—alpha	Hyperparameter for separating regularization loss
—beta	Hyperparameter for focusing regularization loss
—phi_dim	Dimensionality of phi
—split_idx	Index of splits provided by (Pei et al., 2020)

For example, if you want to run on Cora dataset with 0-th split,

python main.py --dataset cora --split_idx 0

Citation

if this work is useful for your research, please cite our paper:

@inproceedings{park2022deformable,
  title={Deformable Graph Convolutional Networks},
  author={Park, Jinyoung and Yoo, Sungdong and Park, Jihwan and Kim, Hyunwoo J},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={36},
  number={7},
  pages={7949--7956},
  year={2022}
}

Acknowledgement

This repo is built upon the following work:

Geom-GCN: Geometric Graph Convolutional Networks. Hongbin Pei, Bingzhe Wei, Kevin Chen-Chuan Chang, Yu Lei, and Bo Yang. ICLR 2020.
Code : https://github.com/graphdml-uiuc-jlu/geom-gcn