Unlink to Unlearn

取消連結以取消學習：簡化 GNN 中的邊緣取消學習

「Unlink to Unlearn：Simplifying Edge Unlearning in GNNs」的程式碼庫。

環境

 torch == 2.1.0
torch_geometric == 2.4.0
numpy == 1.26.2
sklearn == 1.3.2
networkx == 3.0
ogb ==1.3.6
wandb
tqdm

用法

更多訓練參數可以在framework/training_args.py中看到。

訓練

 python train_gnn . py - - dataset Cora

忘記學習

 python delete_gnn . py - - dataset Cora - - df_size 5.0 - - unlearning_model utu 

謝謝

有些程式碼是從 GNNDelete 的程式碼儲存庫中派生的。

引文

 @inproceedings{tan2024unlink,
author = {Tan, Jiajun and Sun, Fei and Qiu, Ruichen and Su, Du and Shen, Huawei},
title = {Unlink to Unlearn: Simplifying Edge Unlearning in GNNs},
year = {2024},
doi = {10.1145/3543873.3587375},
booktitle = {Companion Proceedings of the ACM on Web Conference 2024},
pages = {489–492},
series = {WWW '24},
}