GGNN implementation for reproducing verum results

[4OH4]

Hello there, fascinating work - thank you for sharing your code!

If I may, I would like to ask how you implement the GGNN:

• There is code in Vuld_SySe/graph_network/gnn.py that defines a GatedGraphNeuralNetwork object, but I can't see where this is called (apart from in gnn.main())
• The readme mentions your implementation of Devign in the Running GGNN section, which includes GGNNSum and DevignModel models.

So, to reproduce the results from your paper (ReVeal on Chrome+Debian, verum), do we need to use the GGNN implementation from your Devign repository? And if so, which model should be used? DevignModel is the default.

Thanks in advance for you help

[davidhin]

I assume this is the case. The author has not really been responding to these questions, so I can take a guess. They probably started implementing GNN using something this as reference https://github.com/pcyin/pytorch-gated-graph-neural-network/blob/master/gnn.py. However, they later changed to Deep Graph Library, but left these files in the reproduction package, even though they aren't used.