Generative Modeling on OMNIGLOT binarized (test)
103.03NLLλ-VAE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| λ-VAEdelta=1.01, Architecture=three-layer MLP, K=302026.07 | 103.03 | 0.189 | 1 | 67.7 | |
| VampPriorArchitecture=three-layer MLP, K=302026.07 | 104.75 | 0.192 | — | — | |
| λ-VAEdelta=1.10, Architecture=three-layer MLP, K=302026.07 | 110.54 | 0.203 | 1 | 37.4 | |
| λ-VAEdelta=1.19, Architecture=three-layer MLP, K=302026.07 | 111.14 | 0.204 | 0.9667 | 32.4 | |
| β-VAEbeta=2.0, Architecture=three-layer MLP, K=302026.07 | 119.86 | 0.22 | 0.4333 | 15 | |
| Standard VAElambda=1.0, Architecture=three-layer MLP, K=302026.07 | 123.11 | 0.226 | 0.5667 | 25.6 | |
| β-VAEbeta=0.5, Architecture=three-layer MLP, K=302026.07 | 131.65 | 0.242 | 0.7 | 38.4 |