This repository is the official implementation of Personalized PageRank Graph Attention Networks (ICASSP 2022) (Paper | Poster).

We provide the implementation of PPRGAT in Pytorch. The repository is organised as follows:

• layers/ contains the Graph Attention Network (GAT) layers that are necessary for our experiments;
• models/ contains the implementation of the PPRGAT network (pprgat.py) alongside the benchmark models we used in our experiments;
• utils/ contains the utility functions related to computing the approximate Personalized PageRank (PPR) matrix.

The implementation has been tested under Python 3.9.7 and CUDA 11.2, with the following packages:

• pytorch==1.10.0
• torch-geometric==2.0.4
• numpy==1.22.3
• scipy==1.8.0

Personalized PageRank Graph Attention Networks

@inproceedings{
  choi2022personalized,
  author={Choi, Julie},
  booktitle={ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  title={Personalized Pagerank Graph Attention Networks},
  year={2022},
  volume={},
  number={},
  pages={3578-3582},
  doi={10.1109/ICASSP43922.2022.9746788}
}

MIT