search

FDUDSDE / MAGIC

Codes and data for USENIX Security 24 paper "MAGIC: Detecting Advanced Persistent Threats via Masked Graph Representation Learning"
MIT License
64 stars 10 forks source link

Question about the code of model #16

Closed cmh14 closed 2 months ago

cmh14 commented 2 months ago

Hello sir, I am learning your code for this project and have finished the data-processing part. However, I failed to understand the code of GATConv and GAT classes in /model/gat.py. Especially, the mechanism by which information or parameters are passed in these codes. Would you mind sharing more detailed documentation of these codes or any other related tutorials?

Jimmyokok commented 2 months ago

The GATConv code is simply the official implementation of GATConv of DGL, plus slight modifications on the attention mechanism to include edge features. You will find better documentation at DGL's official API reference. The GAT code implements stacked layers of GATConv, which is modified from GraphMAE's implementation. This part is less interesting and easy to understand after figuring out how GATConv works.

cmh14 commented 2 months ago

Thank you sir! I am reading the dgl document right now. Thank you for your answer.