Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

About Random Seeds #147

Open
L1J1aJun opened this issue Sep 11, 2023 · 2 comments
Open

About Random Seeds #147

L1J1aJun opened this issue Sep 11, 2023 · 2 comments

Comments

@L1J1aJun
Copy link

L1J1aJun commented Sep 11, 2023

Why do I get different results when I set random seeds before and after the dataset? Taking the GCN model as an example, using the CORA dataset after meta attack (ptd_rate=0.05), the model's performance differs greatly when I set random seeds before and after the dataset.

@ChandlerBang
Copy link
Collaborator

I am not exactly sure about the described scenario, but we are supposed to use the same data split before and after attack. This ensures that the attacker uses the correct supervision to learn the perturbation. Otherwise, the attack is very likely to degrade the model performance.

@L1J1aJun
Copy link
Author

我不完全确定所描述的场景,但我们应该在攻击前后使用相同的数据拆分。这可确保攻击者使用正确的监督来学习扰动。否则,攻击很可能会降低模型性能。

感谢你的回复,我发现这个错误发生在使用nettack划分的时候,但是prognn划分不会出现这个问题

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants