This repository contains a PyTorch implementation of "Scalable Graph Neural Networks via Bidirectional Propagation".(http://arxiv.org/abs/2010.15421)

• CUDA 10.1.243
• python 3.6.10
• pytorch 1.4.0
• GCC 5.4.0
• cnpy
• swig-4.0.1

The data folder includes three benchmark datasets(Cora, Citeseer, Pubmed). Other datasets can be downloaded from PPI, Yelp, Amazon2M and Friendster. We also provide code to convert datasets to our format (in convert folder).

make

• To replicate the transductive learning results (Cora, Citeseer, Pubmed), run the following script

sh transductive.sh

• To replicate the inductive learning results (PPI, Yelp, Amazon2M), run the following script

sh inductive.sh

• To replicate the inductive results of Friendster, run the following script

sh friendster.sh

@article{cwdlydw2020gbp,
  title = {Scalable Graph Neural Networks via Bidirectional Propagation},
  author = {Ming Chen, Zhewei Wei, Bolin Ding, Yaliang Li, Ye Yuan, Xiaoyong Du and Ji-Rong Wen},
  year = {2020},
  booktitle = {{NeurIPS}},
}