Table 1 Accuracy (%) of unsupervised multi-source domain adaptation on DigitFive
From: A framework for self-supervised federated domain adaptation
Methods | Target | |||||
|---|---|---|---|---|---|---|
Mnist | Mnist-m | Svhn | Syn | Usps | Avg | |
Upper and lower bounds | ||||||
 Oracle | 99.5 | 95.4 | 92.3 | 98.7 | 99.2 | 97.0 |
 Source-only | 92.3 | 63.7 | 71.5 | 83.4 | 90.71 | 80.3 |
Common multi-source domain adaptation | ||||||
 DAN | 96.3 | 63.8 | 72.5 | 85.4 | 94.2 | 82.4 |
 M3SDA | 98.4 | 72.8 | 81.3 | 89.6 | 96.2 | 87.7 |
 DANN | 97.6 | 71.3 | 63.5 | 85.4 | 92.33 | 82.1 |
Decentralized multi-source domain adaptation | ||||||
 FADA | 92.5 | 64.5 | 63.5 | 82.8 | 91.7 | 79.0 |
 SFDA | 99.10 | 72.31 | 86.02 | 90.37 | 98.06 | 89.17 |