GOURMET v31.5: GNN Enhancement Report (GraphSAGE GPU)
Generated: 2026-06-11T07:40:04.468900+00:00 Device: CUDA (RTX 4060 Ti)
Results
| Metric | Value |
|---|---|
| Previous AUC (MLP CPU) | 0.8296 |
| Previous AUC (MLP GPU) | 0.8326 |
| Previous AUC (v30) | 0.7716 |
| New AUC (GraphSAGE GPU) | 0.9257 |
| Target | > 0.85 |
| Status | ✅ Target reached |
Architecture
- GraphSAGE-style message passing
- 2-layer aggregation (256 → 128 hidden dims)
- Link prediction via concatenated source/target embeddings
- 150 epochs, batch size 512
Training Data
- Positive: 23,525 (observed_in + hierarchical edges)
- Negative: 23,525 (random unconnected pairs)
- Total: 47,050 examples
v31.5 Stream E — GraphSAGE GPU. 2026-06-11T07:40:04.468917+00:00