Table 6.
The first round of experiment results of USC. We also randomly selected 4 users from the USC, namely P5, P6, P7 and P8, and transferred between each pair. In this round of experiments, Target used 60% of the unlabeled data to participate in the training, using the remaining 40% for testing. The bold number is the best result for the experiments in the same row.
| Pairs | Only-Source | MMD | DANN | WD |
|---|---|---|---|---|
| P5~P6 | 0.5965 | 0.6232 | 0.6205 | 0.5775 |
| P6~P5 | 0.5639 | 0.5855 | 0.5318 | 0.6084 |
| P5~P7 | 0.7713 | 0.8063 | 0.7788 | 0.7814 |
| P7~P5 | 0.6400 | 0.7099 | 0.6912 | 0.7292 |
| P5~P9 | 0.5543 | 0.6627 | 0.5429 | 0.5331 |
| P9~P5 | 0.4874 | 0.6787 | 0.5224 | 0.5744 |
| P6~P7 | 0.6625 | 0.6828 | 0.5802 | 0.6507 |
| P7~P6 | 0.5342 | 0.5641 | 0.5580 | 0.6434 |
| P6~P9 | 0.3492 | 0.4483 | 0.4145 | 0.2995 |
| P9~P6 | 0.4370 | 0.6640 | 0.5495 | 0.4818 |
| P7~P9 | 0.4244 | 0.5808 | 0.4994 | 0.4485 |
| P9~P7 | 0.5370 | 0.7174 | 0.5286 | 0.5293 |
| Average | 0.5465 | 0.6436 | 0.5682 | 0.5714 |