5-way 3-shot Image Classification on MiniImageNet 33.3% symmetric label swap noise 54 (test)
56.89AccuracyOracle
Evaluation results
| Method | Links | |
|---|---|---|
| Oracle2022.04 | 56.89 | |
| TraNFS-3Layers=32022.04 | 53.84 | |
| TraNFS-2Layers=22022.04 | 53.35 | |
| Absolute T = 10.0Distance Metric=Absolute, Temperature (T)=10.02022.04 | 50.93 | |
| Absolute T = 25.0Distance Metric=Absolute, Temperature (T)=25.02022.04 | 50.84 | |
| Euclidean T = 10.0Distance Metric=Euclidean, Temperature (T)=10.02022.04 | 50.83 | |
| MedianType=Spatial median prototypes2022.04 | 50.4 | |
| Absolute T = 50.0Distance Metric=Absolute, Temperature (T)=50.02022.04 | 50.06 | |
| Euclidean T = 25.0Distance Metric=Euclidean, Temperature (T)=25.02022.04 | 50.06 | |
| Cosine T = 0.2Distance Metric=Cosine, Temperature (T)=0.22022.04 | 49.63 | |
| Euclidean T = 50.0Distance Metric=Euclidean, Temperature (T)=50.02022.04 | 49.51 | |
| Cosine T = 1.0Distance Metric=Cosine, Temperature (T)=1.02022.04 | 49.2 | |
| Cosine T = 0.5Distance Metric=Cosine, Temperature (T)=0.52022.04 | 49.15 | |
| Cosine T = 2.0Distance Metric=Cosine, Temperature (T)=2.02022.04 | 49.05 | |
| Cosine T = 5.0Distance Metric=Cosine, Temperature (T)=5.02022.04 | 48.96 | |
| Vanilla ProtoNet2022.04 | 48.78 | |
| RNNP2022.04 | 48.76 | |
| Linear Classifier2022.04 | 46.06 | |
| Matching Networks2022.04 | 44.92 | |
| MAML2022.04 | 40.41 | |
| Nearest kk=52022.04 | 40.05 | |
| Nearest kk=12022.04 | 39.92 | |
| Nearest kk=32022.04 | 38.76 |