Architecture Performance Prediction on NAS-Bench-101 (test)
0.903Kendall's TauGATES
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GATESProportion of training samples=100%2020.04 | 0.903 | — | |
| GATESProportion of training samples=50%2020.04 | 0.9001 | — | |
| MLPProportion of training samples=100%2020.04 | 0.8955 | — | |
| GCN (w.o. global node)Proportion of training samples=100%2020.04 | 0.8952 | — | |
| GCN (global node)Proportion of training samples=100%2020.04 | 0.895 | — | |
| LSTMProportion of training samples=100%2020.04 | 0.8931 | — | |
| GATESProportion of training samples=10%2020.04 | 0.8922 | — | |
| GCN (global node)Proportion of training samples=50%2020.04 | 0.8918 | — | |
| GCN (w.o. global node)Proportion of training samples=50%2020.04 | 0.891 | — | |
| MLPProportion of training samples=50%2020.04 | 0.8893 | — | |
| LSTMProportion of training samples=50%2020.04 | 0.8859 | — | |
| GATESProportion of training samples=5%2020.04 | 0.8841 | — | |
| GCN (global node)Proportion of training samples=10%2020.04 | 0.8747 | — | |
| GCN (w.o. global node)Proportion of training samples=10%2020.04 | 0.8721 | — | |
| MLPProportion of training samples=10%2020.04 | 0.8718 | — | |
| GCN (global node)Proportion of training samples=5%2020.04 | 0.8641 | — | |
| GCN (w.o. global node)Proportion of training samples=5%2020.04 | 0.8626 | — | |
| GATESProportion of training samples=1%2020.04 | 0.8594 | — | |
| MLPProportion of training samples=5%2020.04 | 0.8592 | — | |
| LSTMProportion of training samples=10%2020.04 | 0.8576 | — | |
| LSTMProportion of training samples=5%2020.04 | 0.844 | — | |
| GATESProportion of training samples=0.5%2020.04 | 0.8434 | — | |
| GCN (global node)Proportion of training samples=1%2020.04 | 0.8277 | — | |
| GCN (w.o. global node)Proportion of training samples=1%2020.04 | 0.8243 | — | |
| GCN (global node)Proportion of training samples=0.5%2020.04 | 0.7915 | — | |
| GATESProportion of training samples=0.1%2020.04 | 0.7789 | — | |
| LSTMProportion of training samples=1%2020.04 | 0.7747 | — | |
| GATESProportion of training samples=0.05%2020.04 | 0.7634 | — | |
| MLPProportion of training samples=1%2020.04 | 0.7312 | — | |
| LSTMProportion of training samples=0.5%2020.04 | 0.7112 | — | |
| GCN (w.o. global node)Proportion of training samples=0.5%2020.04 | 0.6963 | — | |
| CARL2026.05 | 0.68 | 93.93 | |
| MLPProportion of training samples=0.5%2020.04 | 0.6463 | — | |
| NP2026.05 | 0.62 | 93.74 | |
| LSTMProportion of training samples=0.1%2020.04 | 0.5993 | — | |
| CAP2026.05 | 0.59 | 93.5 | |
| GCN (global node)Proportion of training samples=0.1%2020.04 | 0.579 | — | |
| ConvNP2026.05 | 0.57 | 94.09 | |
| LSTMProportion of training samples=0.05%2020.04 | 0.5509 | — | |
| GCN (global node)Proportion of training samples=0.05%2020.04 | 0.5343 | — | |
| RF2026.05 | 0.53 | 93.71 | |
| MLPProportion of training samples=0.1%2020.04 | 0.5272 | — | |
| XGB2026.05 | 0.49 | 93.59 | |
| RR2026.05 | 0.47 | 93.88 | |
| GCN (w.o. global node)Proportion of training samples=0.1%2020.04 | 0.4628 | — | |
| LR2026.05 | 0.46 | 93.8 | |
| GCN (w.o. global node)Proportion of training samples=0.05%2020.04 | 0.3992 | — | |
| MLPProportion of training samples=0.05%2020.04 | 0.3971 | — | |
| MLP2026.05 | 0.31 | 93.47 | |
| RE2026.05 | — | 93.68 | |
| RS2026.05 | — | 93.55 |