Few-shot classification on MiniImageNet 5-way 10-shot (test)
71.78AccuracyOracle
Evaluation results
| Method | Links | |
|---|---|---|
| OracleNoise Proportion=20%2022.04 | 71.78 | |
| OracleNoise Proportion=30%2022.04 | 70.82 | |
| TraNFSn=3, Noise Proportion=20%2022.04 | 70.38 | |
| TraNFSn=2, Noise Proportion=20%2022.04 | 70.13 | |
| OracleNoise Proportion=40%2022.04 | 69.27 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=25, Noise Proportion=20%2022.04 | 69.12 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=50, Noise Proportion=20%2022.04 | 69.07 | |
| Spatial median prototypesAggregation=Median, Noise Proportion=20%2022.04 | 69.04 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=25, Noise Proportion=20%2022.04 | 69 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=10, Noise Proportion=20%2022.04 | 68.91 | |
| RNNPNoise Proportion=20%2022.04 | 68.89 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=50, Noise Proportion=20%2022.04 | 68.86 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.2, Noise Proportion=20%2022.04 | 68.82 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.5, Noise Proportion=20%2022.04 | 68.58 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=5, Noise Proportion=20%2022.04 | 68.5 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=1, Noise Proportion=20%2022.04 | 68.48 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=2, Noise Proportion=20%2022.04 | 68.42 | |
| Vanilla ProtoNetNoise Proportion=20%2022.04 | 68.34 | |
| TraNFSn=3, Noise Proportion=30%2022.04 | 67.03 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=10, Noise Proportion=20%2022.04 | 66.94 | |
| TraNFSn=2, Noise Proportion=30%2022.04 | 66.2 | |
| RNNPNoise Proportion=30%2022.04 | 63.86 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=25, Noise Proportion=30%2022.04 | 63.75 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=10, Noise Proportion=30%2022.04 | 63.73 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=50, Noise Proportion=30%2022.04 | 63.62 | |
| Linear ClassifierNoise Proportion=20%2022.04 | 63.54 | |
| Spatial median prototypesAggregation=Median, Noise Proportion=30%2022.04 | 63.5 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=25, Noise Proportion=30%2022.04 | 63.48 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=50, Noise Proportion=30%2022.04 | 63.2 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.2, Noise Proportion=30%2022.04 | 63.14 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.5, Noise Proportion=30%2022.04 | 62.76 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=5, Noise Proportion=30%2022.04 | 62.72 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=2, Noise Proportion=30%2022.04 | 62.63 | |
| Vanilla ProtoNetNoise Proportion=30%2022.04 | 62.59 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=1, Noise Proportion=30%2022.04 | 62.59 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=10, Noise Proportion=30%2022.04 | 61.82 | |
| TraNFSn=3, Noise Proportion=40%2022.04 | 58.94 | |
| Matching NetworksNoise Proportion=20%2022.04 | 57.74 | |
| TraNFSn=2, Noise Proportion=40%2022.04 | 56.97 | |
| Linear ClassifierNoise Proportion=30%2022.04 | 56.7 | |
| MAMLNoise Proportion=20%2022.04 | 55.05 | |
| RNNPNoise Proportion=40%2022.04 | 54.06 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=25, Noise Proportion=40%2022.04 | 53.88 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=10, Noise Proportion=40%2022.04 | 53.81 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=50, Noise Proportion=40%2022.04 | 53.78 | |
| Spatial median prototypesAggregation=Median, Noise Proportion=40%2022.04 | 53.61 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=25, Noise Proportion=40%2022.04 | 53.58 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.2, Noise Proportion=40%2022.04 | 53.27 | |
| Similarity weighted prototypesMetric=Euclidean, Temperature (T) value=50, Noise Proportion=40%2022.04 | 53.24 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=0.5, Noise Proportion=40%2022.04 | 52.9 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=2, Noise Proportion=40%2022.04 | 52.87 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=1, Noise Proportion=40%2022.04 | 52.86 | |
| Matching NetworksNoise Proportion=30%2022.04 | 52.8 | |
| Similarity weighted prototypesMetric=Cosine, Temperature (T) value=5, Noise Proportion=40%2022.04 | 52.76 | |
| Vanilla ProtoNetNoise Proportion=40%2022.04 | 52.73 | |
| Similarity weighted prototypesMetric=Absolute, Temperature (T) value=10, Noise Proportion=40%2022.04 | 52.28 | |
| Nearest Neighborsk=5, Noise Proportion=20%2022.04 | 49.88 | |
| Linear ClassifierNoise Proportion=40%2022.04 | 47.85 | |
| Nearest Neighborsk=3, Noise Proportion=20%2022.04 | 46.96 | |
| Matching NetworksNoise Proportion=40%2022.04 | 45.37 | |
| Nearest Neighborsk=5, Noise Proportion=30%2022.04 | 45.21 | |
| Nearest Neighborsk=1, Noise Proportion=20%2022.04 | 44.85 | |
| Nearest Neighborsk=3, Noise Proportion=30%2022.04 | 42.31 | |
| MAMLNoise Proportion=30%2022.04 | 41.95 | |
| MAMLNoise Proportion=40%2022.04 | 41.83 | |
| Nearest Neighborsk=1, Noise Proportion=30%2022.04 | 40.8 | |
| Nearest Neighborsk=5, Noise Proportion=40%2022.04 | 39.47 | |
| Nearest Neighborsk=3, Noise Proportion=40%2022.04 | 37.32 | |
| Nearest Neighborsk=1, Noise Proportion=40%2022.04 | 36.58 |