This repository provides the reference implementation of CorePPR for a single machine in TensorFlow 1.
CorePPR is a scalable model that utilises a learnable convex combination of the approximate personalised PageRank and the CoreRank to diffuse multi-hop neighbourhood information in GNNs. It was proposed in our paper
Improving Graph Neural Networks at Scale: Combining Approximate PageRank and CoreRank
by Ariel R. Ramos Vela, Johannes F. Lutzeyer, Anastasios Giovanidis, Michalis Vazirgiannis
Accepted at the "NeurIPS 2022 New Frontiers in Graph Learning Workshop (NeurIPS GLFrontiers 2022)"
You can install the repository using pip install -e .. However, installing the requirements like this will result in TensorFlow using CUDA 10.0, which contains a bug that affects PPRGo. We recommend importing the Anaconda environment saved in environment.yaml instead, which provides the correct TensorFlow and CUDA versions.
To see for yourself how CorePPR performs on a large dataset, we have included a (demo.ipynb) notebook that trains and generates predictions for the datasets described in the paper.
Please contact ariel.ramosvela@ip-paris.fr or johannes.lutzeyer@polytechnique.edu. if you have any question.
Please cite our paper if you use the model or this code in your own work:
@misc{https://doi.org/10.48550/arxiv.2211.04248,
doi = {10.48550/ARXIV.2211.04248},
url = {https://arxiv.org/abs/2211.04248},
author = {Vela, Ariel R. Ramos and Lutzeyer, Johannes F. and Giovanidis, Anastasios and Vazirgiannis, Michalis},
keywords = {Machine Learning (cs.LG), Social and Information Networks (cs.SI), Machine Learning (stat.ML), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Improving Graph Neural Networks at Scale: Combining Approximate PageRank and CoreRank},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution Non Commercial Share Alike 4.0 International}
}