Few-shot classification on Omniglot 20-way 1-shot (test)
99.12AccuracyMC2
Evaluation results
| Method | Links | |
|---|---|---|
| MC2Ensemble=3 model2019.02 | 99.12 | |
| MC22019.02 | 97.86 | |
| MAML++2019.02 | 97.65 | |
| SNAIL2019.02 | 97.64 | |
| MC12019.02 | 97.6 | |
| GNN2019.02 | 97.4 | |
| Meta-SGD2019.02 | 95.93 | |
| MAML2019.02 | 95.8 | |
| ProtoNetsClustering=N/A, Supervised=true2020.04 | 95.31 | |
| MAMLClustering=N/A, Supervised=true2020.04 | 84.6 | |
| Units-NTSLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=NTS2022.03 | 83.56 | |
| Units-MLLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=ML2022.03 | 83.45 | |
| Versa TSLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=TS2022.03 | 81.85 | |
| Versa NTSLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=NTS2022.03 | 80.52 | |
| Versa TSLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=TS2022.03 | 79.93 | |
| Units-MLLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=ML2022.03 | 79.17 | |
| Units-MLLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=ML2022.03 | 78.6 | |
| Versa NTSLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=NTS2022.03 | 78.33 | |
| ULDA-ProtoNetsClustering=N/A2020.04 | 78.05 | |
| ULDA-MetaOptNetClustering=N/A2020.04 | 76.32 | |
| Units-NTSLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=NTS2022.03 | 75.88 | |
| UMTRAClustering=N/A2020.04 | 74.25 | |
| MAMLLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=Random2022.03 | 73.91 | |
| Units-NTSLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=NTS2022.03 | 73.35 | |
| UA-MetaSGD TSLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=TS2022.03 | 72.51 | |
| MAMLLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=Random2022.03 | 72.41 | |
| Versa TSLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=TS2022.03 | 72.15 | |
| UA-MetaSGD NTSLabeling budget (Number of Tasks)=10000 Tasks, Task selection protocol=NTS2022.03 | 71.44 | |
| UA-MetaSGD NTSLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=NTS2022.03 | 71.34 | |
| UA-MetaSGD TSLabeling budget (Number of Tasks)=8000 Tasks, Task selection protocol=TS2022.03 | 71.07 | |
| Versa NTSLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=NTS2022.03 | 70.24 | |
| AAL-MAML++Clustering=N/A2020.04 | 70.21 | |
| AAL-ProtoNetsClustering=N/A2020.04 | 68.79 | |
| MAMLLabeling budget (Number of Tasks)=1000 Tasks, Task selection protocol=Random2022.03 | 65.24 | |
| UA-MetaSGD TSLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=TS2022.03 | 62.38 | |
| UA-MetaSGD NTSLabeling budget (Number of Tasks)=4000 Tasks, Task selection protocol=NTS2022.03 | 61.94 | |
| CACTUS-MAMLClustering=ACAI2020.04 | 48.09 | |
| CACTUS-ProtoNetsClustering=ACAI2020.04 | 47.75 | |
| linear classifierClustering=DeepCluster2020.04 | 27.8 | |
| knn-nearest neighborsClustering=DeepCluster2020.04 | 27.37 | |
| Training from scratchClustering=N/A2020.04 | 24.91 | |
| cluster matchingClustering=DeepCluster2020.04 | 21.54 | |
| MLP with dropoutClustering=DeepCluster2020.04 | 19.92 |