thudm/grand-plus

Code and dataset for paper "GRAND+: Scalable Graph Random Neural Networks"

GRAND+

This is a PyTorch implementation of GRAND+ for scalable graph-based semi-supervised learning:

GRAND+: Scalable Graph Random Neural Networks

You may be also interested in the predecessor of this work: Graph Random Neural Network for Semi-Supervised Learning on Graphs [github repo].

Datasets

This repo contains Cora, Citeseer and Pubmed datasets under the path dataset/citation/. The other datasets used in the paper (including AMiner-CS, Reddit, Amazon2M and MAG-Scholar-C) can be downloaded from Google Drive or Tsinghua Cloud. To run model on these datasets, you should download the corresponding zip file, uncompress it and put it under dataset/.

You can directly download the zip file of each dataset with the following scripts:

• Download datasets from Google Drive

pip install gdown
gdown --id 1G9Wn1OaqMYpkNmbOESYUFrDgzo0Be0-L -O dataset/aminer.zip
gdown --id 1KauMd-AJXyD6KQQnf4vySjRZEOgWQYvx -O dataset/reddit.zip
gdown --id 1uItY1AGywFv4nSSFpqBaTEUoDn3w414B -O dataset/Amazon2M.zip
gdown --id 1VKHFQfRXkkVShE6d4hA9dImXZalz49qa -O dataset/mag_scholar_c.npz

• Download datasets from Tsinghua Cloud

python scripts/download.py --url https://cloud.tsinghua.edu.cn/f/629a605e453b40fc9a93/?dl=1 --path dataset --fname aminer.zip
python scripts/download.py --url https://cloud.tsinghua.edu.cn/f/384be92876ed4127aa3c/?dl=1 --path dataset --fname reddit.zip
python scripts/download.py --url https://cloud.tsinghua.edu.cn/f/7c867cef16214fe1a30b/?dl=1 --path dataset --fname Amazon2M.zip
python scripts/download.py --url https://cloud.tsinghua.edu.cn/f/5e5c9d8833a143d5abb4/?dl=1 --path dataset --fname mag_scholar_c.npz

Requirements

• g++ 7.5.0
• pybind11
• networkx 2.5
• numpy 1.19.2
• scikit_learn 1.0.2
• scipy 1.5.2
• torch 1.8.1 (cuda 10.2)
• torch_scatter 2.0.6

Compilation

make clean && make

Running the code

sh scripts/run_<dataset>.sh <runs> <cuda_id> <propagation matrix [ppr, avg, single]>

Example:

• Running model on Pubmed for 10 runs with personalized pagerank matrix: sh scripts/run_pubmed.sh 10 <cuda_id> ppr

Cite

If you find this work is helpful to your research, please consider citing our paper:

@inproceedings{feng2022grand+,
  title={GRAND+: Scalable Graph Random Neural Networks},
  author={Feng, Wenzheng and Dong, Yuxiao and Huang, Tinglin and Yin, Ziqi and Cheng, Xu and Kharlamov, Evgeny and Tang, Jie},
  booktitle={Proceedings of the ACM Web Conference 2022 (WWW’22)},
  year={2022}
}

grand-plus's Issues

Suggest to loosen the dependency on requests

Hi, your project GRAND-plus(commit id: 5b960cf) requires "requests==2.24.0" in its dependency. After analyzing the source code, we found that the following versions of requests can also be suitable, i.e., requests 2.22.0, 2.23.0, since all functions that you directly (2 APIs: requests.sessions.Session.init, requests.exceptions.ConnectionError.init) or indirectly (propagate to 13 requests's internal APIs and 5 outsider APIs) used from the package have not been changed in these versions, thus not affecting your usage.

Therefore, we believe that it is quite safe to loose your dependency on requests from "requests==2.24.0" to "requests>=2.22.0,<=2.24.0". This will improve the applicability of GRAND-plus and reduce the possibility of any further dependency conflict with other projects.

May I pull a request to further loosen the dependency on requests?

By the way, could you please tell us whether such an automatic tool for dependency analysis may be potentially helpful for maintaining dependencies easier during your development?