spektral
spektral copied to clipboard
[Feature Request] GraphSAGE model support for edge features
Hi,
As I understand, the current GraphSAGE model takes into account the node level features for model training. Is there any example / support for GraphSAGE model which also supports edge features for edge classification tasks. Or a similar variant that can be used as an inductive learning approach ? This can be very useful in some use cases relating to detecting malicious flows in the network data etc.
Thanks!
Here is the paper which references such model as E-GraphSage https://arxiv.org/pdf/2103.16329.pdf
Hi,
This is in the pipeline but I can't give you a timeline for the development because I've been super busy and I barely have the time to manage issues/bugs as it is. I'll change the issue to a Feature Request so that I'll work on it when I have time or someone else might give it a shot.
Cheers
Sure, understood and sounds good, thanks. If you've some pointers in mind, I can also help contribute towards this. This looks to be an amazing framework to build bunch of graph NN use cases.