Disentangled Representation Learning on Noisy Swiss Roll ρ = 0.3
0.502Mutual Information I(z; w)CCVAE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CCVAENoise level (ρ)=0.32025.06 | 0.502 | 3.302 | 2.21 | |
| DIVANoise level (ρ)=0.32025.06 | 0.336 | 3.302 | 1.88 | |
| CSVAEAdversarial training=false, Noise level (ρ)=0.32025.06 | 0.047 | 3.303 | 23.88 | |
| CSVAEAdversarial training=true, Noise level (ρ)=0.32025.06 | 0.031 | 3.302 | 17.99 | |
| HCSVAEAdversarial training=false, Noise level (ρ)=0.32025.06 | 0.024 | 3.302 | 30 | |
| DisCoVRNoise level (ρ)=0.32025.06 | 0.005 | 3.302 | 1.14 | |
| HCSVAEAdversarial training=true, Noise level (ρ)=0.32025.06 | 0.002 | 3.302 | 30 |