20-way Few-shot Classification on Omniglot (test)
98.8251-shot AccuracyDCN4
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DCN4layers=42019.09 | 98.825 | 99.505 | |
| Support nets6layers=62019.09 | 98.58 | 99.45 | |
| Support nets4layers=42019.09 | 97.79 | 99.27 | |
| SNAIL2019.09 | 97.64 | 99.36 | |
| Relation nets2019.09 | 97.6 | 99.1 | |
| MetaNet2017.12 | 97 | — | |
| Prototypical NetworksDistance=Euclidean, Fine-tune=No2017.03 | 96 | 98.9 | |
| Prototypical nets2019.09 | 96 | 98.9 | |
| Meta-SGD2019.09 | 95.93 | 98.97 | |
| MAML2019.09 | 95.8 | 98.9 | |
| Matching NetworksDistance=Cosine, Fine-tune=No2017.03 | 93.8 | 98.5 | |
| Matching Nets2017.12 | 93.8 | 98.5 | |
| Matching nets2019.09 | 93.8 | 98.5 | |
| Matching NetworksDistance=Cosine, Fine-tune=Yes2017.03 | 93.5 | 98.7 | |
| Neural StatisticianFine-tune=No2017.03 | 93.2 | 98.1 | |
| Neural statistician2019.09 | 93.2 | 98.1 | |
| Siamese Net2017.12 | 88.2 | 97 | |
| Siamese nets2019.09 | 88.1 | 97 |