GCN · internal_only
Standard SplitFeb 17, 20268137724c215e436080a432fb2d39fc8f
Description
Train GCN on the internal_only dataset to compare GNN architectures. Same hyperparameters and data split as GIN experiments.
Conclusion
GCN performs well (96.3% accuracy) but below GIN (98.2%). Recall drops to 93.8% — GCN misses some malware samples. GIN remains the stronger architecture on this dataset.
Test Metrics
Accuracy
96.3%
F1 Macro
95.6%
F1 Malware
93.8%
Precision
93.8%
Recall
93.8%
AUROC
98.4%
Best Val Loss
0.1720
Training Time
1057.2000s
Confusion Matrix
| Pred Benign | Pred Malware | |
|---|---|---|
| Actual Benign | 74 | 2 |
| Actual Malware | 2 | 30 |
Configuration
| Hidden Dim | 128 |
| Num Layers | 3 |
| Dropout | 0.5 |
| Batch Size | 4 |
| Learning Rate | 0.001 |
| Weight Decay | 0.0001 |
| Max Epochs | 200 |
| ES Patience | 20 |
| ES Min Epochs | 100 |
| LR Patience | 10 |
| LR Factor | 0.5 |
| Mixed Precision | Yes |
| Random Seed | 42 |
| Epochs Trained | 108 |