FairU

A study of Metric and Framework Improving Fairness-utility Trade-off in Link Prediction

Overall Framework of FairU

Setups

Datasets

The dataset name must be specified in the "--dataset" argument

• Citeseer (using latest 1/64 fraction due to the amount of full dataset)
• Cora
• Facebook

After downloaded the datasets, you can put them in the folder data/ like the following.

$ tree
.
├── Citeseer
│   ├── ind.citeseer.allx
│   ├── ...
│   └── ind.citeseer.y
├── Cora
│   ├── cora.cites
│   └── cora.contents
└── Facebook
    ├── facebook
    │    ├── 0.circles
    │    ├── ...    
    │    └── 3980.featnames
    ├── facebook_combinded.txt
    └── readme-Ego.txt

Train and Test

# Citeseer
python main.py --n_epochs 200 --device cuda:2 --adv True --alpha 1 --beta 0 --dataset citeseer --fairdrop_term 30

# Cora
python main.py --n_epochs 200 --device cuda:0 --adv True --alpha 1 --beta 0  --dataset cora --fairdrop_term 10

# Facebook
python main.py --n_epochs 200 --device cuda:1 --adv True --alpha 0.8 --beta 0 --dataset facebook --fairdrop_term 10

Citation

Please cite our paper if you use the code:

@article{yang2023fairu,
  title={A Study of Metric and Framework Improving Fairness-utility Trade-off in Link Prediction},
  author={Heeyoon Yang, YongHoon Kang, Gahyung Kim, Jiyoung Lim, SuHyun Yoon, Ho Seung Kim, Jee-Hyong Lee},
  journal={Journal of KIISE},
  year={2023},
  doi={10.5626/JOK.2023.50.2.179}
}