Fig. 5
From: Bin2vec: learning representations of binary executable programs for security tasks
Comparison of GCN models with different numbers of layers (A), and different sizes of each layer (C). (A) demonstrates the accuracy of GCN models with different numbers of layers - from 1 to 8 - on the validation set of Task 1; the size of each layer is 64. (B) shows the time in seconds that models from (A) take to train on 100 samples. (C) demonstrates the accuracy of GCN models with 3 layers, but different sizes of layers - from 32 to 256. (D) shows the time in seconds that models from (C) take to train on 100 samples