Jerry-jwz/DisC

NeurIPS2022-Debiasing Graph Neural Networks via Learning Disentangled Causal Substructure

DisC

Source code for "NeurIPS2022-Debiasing Graph Neural Networks via Learning Disentangled Causal Substructure"

paper: https://arxiv.org/pdf/2209.14107.pdf

                                                         The framework of DisC

Contact

Shaohua Fan, Email:fanshaohua@bupt.edu.cn

Datasets

Datasets used for Table 1: https://drive.google.com/file/d/1pv_cFKYJxXpT4qJ6jgvNn17MIovZUrhA/view?usp=sharing

Unseen test set for Table 2: https://drive.google.com/file/d/1EBbFh8HDYjO4XpNctPX6U4iVYkHy6qSK/view?usp=sharing # f[0] is the unbiased test set

Requirements

pip -r requirements.txt

Running the model

DisC_GCN

python Disc_gcn_run.py

DisC_Gin

python Disc_gin_run.py

DisC_Gin

python Disc_gcnii_run.py

Reference

@inproceedings{

author = {Shaohua Fan, Xiao Wang, Yanhu Mo, Chuan Shi, Jian Tang},

title = {Debiasing Graph Neural Networks via Learning Disentangled Causal Substructure},

booktitle = {NeurIPS},

year = {2022} }