Classification on Rotating MNIST
0.879AccuracyGradual ST
Evaluation results
| Method | Links | |
|---|---|---|
| Gradual STtraining_strategy=Sequential self-training across intermediate domains, backbone=3-layer CNN, dropout=0.5, batchnorm=Last layer2020.02 | 0.879 | |
| All STtraining_strategy=Self-train on pooled unlabeled data, backbone=3-layer CNN, dropout=0.5, batchnorm=Last layer2020.02 | 0.38 | |
| Target STtraining_strategy=Self-train on target, backbone=3-layer CNN, dropout=0.5, batchnorm=Last layer2020.02 | 0.33 | |
| Sourcetraining_strategy=Supervised on source, backbone=3-layer CNN, dropout=0.5, batchnorm=Last layer2020.02 | 0.319 |