5-way Classification on CIFAR100 Fewshot (test)
45.1Accuracy (1-shot)MTL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MTLLearning strategy=SS [Φ; θ], Meta-batch configuration=HT meta-batch, Feature extractor=ResNet-12 (pre)2018.12 | 45.1 | 57.6 | 63.4 | |
| MTLLearning strategy=SS [Φ; θ], Meta-batch configuration=meta-batch, Feature extractor=ResNet-12 (pre)2018.12 | 43.6 | 55.4 | 62.4 | |
| MAML deep, HTLearning strategy=FT [Θ; θ], Meta-batch configuration=HT meta-batch, Feature extractor=ResNet-12 (pre)2018.12 | 41.8 | 55.1 | 61.9 | |
| TADAMFeature extractor=ResNet-12 (pre), Additional layers=72 fully connected layers2018.12 | 40.1 | 56.1 | 61.6 | |
| MAML, HTLearning strategy=FT [Θ; θ], Meta-batch configuration=HT meta-batch, Feature extractor=4 CONV2018.12 | 39.9 | 51.7 | 57.2 | |
| MAMLFeature extractor=4 CONV, Implementation=authors implementation from public code2018.12 | 38.1 | 50.4 | 56.2 |