Table 3:
Prediction Performance
| Dataset | Proposed | VAE | β-VAE | Factor-VAE | TCVAE | DSD | |
|---|---|---|---|---|---|---|---|
| κ | MNIST | .969 | .494 | .326 | .704 | .260 | .030 |
| Fashion-MNIST | .857 | .389 | .460 | .613 | .482 | .003 | |
|
| |||||||
| R 2 | Yale Faces | .968 | .397 | .760 | .699 | .692 | N/A |
| 3D chairs | .912 | .155 | .357 | .224 | .196 | N/A | |
| 3D cars | .584 | .391 | .418 | .177 | .110 | N/A | |
DSD is designed for analyzing binary similarities, and cannot analyze real-valued similarities.