Accuracy Prediction on NAS-Bench-101 (depth=2~7, Test Samples=100)
0.801Kendall's TauNAR-Former
Evaluation results
| Method | Links | |
|---|---|---|
| NAR-FormerBackbone=Transformer, Training Samples=0.1% (424)2022.11 | 0.801 | |
| TNASP + SEBackbone=Transformer, Training Samples=0.1% (424)2022.11 | 0.754 | |
| TNASPBackbone=Transformer, Training Samples=0.1% (424)2022.11 | 0.752 | |
| CTNASBackbone=GCN, Training Samples=0.1% (424)2022.11 | 0.751 | |
| NAO+SEBackbone=LSTM, Training Samples=0.1% (424)2022.11 | 0.732 | |
| NP + SEBackbone=GCN, Training Samples=0.1% (424)2022.11 | 0.713 | |
| NPBackbone=GCN, Training Samples=0.1% (424)2022.11 | 0.71 | |
| NAOBackbone=LSTM, Training Samples=0.1% (424)2022.11 | 0.704 | |
| ReNASBackbone=CNN, Training Samples=0.1% (424)2022.11 | 0.634 |