Subgraph Federated Unlearning

This study is the subgraph federated unlearning for multi-clients

Environment settings and libraries we used in our experiments

This project is tested under the following environment settings:

• torch>=1.9.0
• torchvision>=0.10.0
• numpy>=1.23.2
• dgl>=0.9.1
• networkx>=2.4
• hdbscan==0.8.28
• joblib==1.1.0

Subgraph Federated Unlearning

Unlearning

python FedUnlearnGNN.py    --dataset Cora  --is_iid iid   --num_workers 10  --num_malicious 1    --trigger_type renyi-u     --overlapping_rate 0.1    --device_id 0  --poisoning_intensity 0.1  --trigger_size 5  --epochs 1 --rounds 400  --ratio_training 0.4 --sample_method str_sample --subgraph_sample_rate 0.3 --k_hop 2