Data

Tools

Indexes

Applications

Experiments

Open Source Science

https://github.com/cgcl-codes/cross-links-bias

Implementation for NeurIPS 2023 paper: Cross-links Matter for Link Prediction: Rethinking the Debiased GNN from a Data Perspective

https://github.com/cgcl-codes/cross-links-bias

Science Score: 13.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ○
    .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (13.9%) to scientific vocabulary
Last synced: 6 months ago · JSON representation

Repository

Implementation for NeurIPS 2023 paper: Cross-links Matter for Link Prediction: Rethinking the Debiased GNN from a Data Perspective

Basic Info
  • Host: GitHub
  • Owner: CGCL-codes
  • Language: Python
  • Default Branch: main
  • Size: 25.4 KB
Statistics
  • Stars: 2
  • Watchers: 2
  • Forks: 1
  • Open Issues: 1
  • Releases: 0
Created over 2 years ago · Last pushed about 1 year ago
Metadata Files
Readme

README.md

Note

This repository includes the implementation for our NeurIPS 2023 paper: Cross-links Matter for Link Prediction: Rethinking the Debiased GNN from a Data Perspective.

Environments

Python 3.7.6

Packages: dgl_cu102==0.9.1.post1 numpy==1.19.2 python_louvain==0.15 networkx==2.5 tqdm==4.62.3 torch==1.12.1+cu102 community==1.0.0b1 dgl==1.1.0 PyYAML==6.0 community is an essential package to deploy the Louvain algorithm used in our work.

Run the following code to install all required packages. ```

pip install -r requirements.txt `` [!NOTE] We notice that nowdgl_cu102==0.9.1.post1` can not be installed by conda/pip directly. One can refer to the previous packages and download the corresponding package before installation.

Datasets & Processed files

  • Due to size limitation, the processed files and datasets are stored in google drive. The datasets include Epinions, DBLP and LastFM.
  • Each dataset directory contains the following processed files:
    • graph.pkl: DGLGraph object for storing the graph structure.
    • split_edge.pkl: Splitted training samples, validation samples and test samples.
    • louvain_dataset.pkl: Detected community memberships through Louvain algorithm.
    • Other processed files for running PPRGo and UltraGCN, such as constrainmat.pkl, iitopk_neighbors.np.pkl.

Run the codes

All arguments are properly set in advance in the script files for reproducing our results.

Here we take GraphSAGE and GAT as examples.

```

bash script/rungraphsagee2e.sh bash script/rungate2e.sh ```

BibTeX

If you like our work and use the model for your research, please cite our work as follows.

bibtex @inproceedings{luo2023cross-links, author = {Luo, Zihan and Huang, Hong and Lian, Jianxun and Song, Xiran and Xie, Xing and Jin, Hai}, title = {Cross-links Matter for Link Prediction: Rethinking the Debiased GNN from a Data Perspective}, booktitle={Thirty-seventh Conference on Neural Information Processing Systems}, year = {2023}, month = {October}, url = {https://www.microsoft.com/en-us/research/publication/cross-links-matter-for-link-prediction-rethinking-the-debiased-gnn-from-a-data-perspective/}, }

Owner

  • Name: CGCL-codes
  • Login: CGCL-codes
  • Kind: organization

CGCL/SCTS/BDTS Lab

GitHub Events

Total
  • Issues event: 1
  • Watch event: 1
  • Issue comment event: 2
  • Push event: 1
  • Fork event: 2
Last Year
  • Issues event: 1
  • Watch event: 1
  • Issue comment event: 2
  • Push event: 1
  • Fork event: 2