edisonleeeee / guard

[CIKM 2023] GUARD: Graph Universal Adversarial Defense

Hi there, I'm EdisonLeeeee! 👋

- 😎 About me

• Sun Yat-sen University, China
• Major in Software Engineering
• Learning Python, Machine Learning
• TensorFlow and PyTorch Enthusiast

- Research on

• Graph Representation Learning
• Trustworthy Graph Learning
• Graph Self-supervised Learning

- Project

• GraphGallery: Graph Gallery for benchmarking graph neural networks
• GreatX (ongoing): Graph reliability toolbox based on PyTorch Geometric
• Mooon (ongoing): Graph data augmentation library based on PyTorch Geometric

guard's Issues

关于模型在计算I时使用的图结构

您好，我在阅读您的文章时发现一个问题，当使用clean graph的degree作为模型的输入时，获得的defense效果要比使用attacked graph的degree作为模型输入好上不少，您是否也注意到过这个问题呢，对于这个结果您怎么看，只是想和您探讨一下，期待您的回复！

No module named 'torch_geometric.nn.dense.linear'

Hi, I do not see any requirement to install torch geometric. So when I ran your demo.ipynb, it caused an error

'No module named 'torch_geometric.nn.dense.linear'

What is the version of torch geometric you used for this repo? I'm using torch geometric 1.7.0

Thanks