Generative Modeling on Dynamically binarized MNIST (test)
-97.62NELBODouble CV
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Double CVNumber of layers=Two layers2022.02 | -97.62 | — | — | |
| DisARMNumber of layers=Two layers2022.02 | -97.56 | — | — | |
| RODEONumber of layers=Two layers2022.02 | -97.43 | — | — | |
| DisARMNumber of layers=Three layers2022.02 | -94.12 | — | — | |
| Double CVNumber of layers=Three layers2022.02 | -93.71 | — | — | |
| DisARMNumber of layers=Four layers2022.02 | -93.56 | — | — | |
| RODEONumber of layers=Three layers2022.02 | -93.52 | — | — | |
| Double CVNumber of layers=Four layers2022.02 | -93.28 | — | — | |
| RODEONumber of layers=Four layers2022.02 | -93.27 | — | — | |
| LSGM2021.06 | 78.47 | — | — | |
| NVAE2021.06 | 79.56 | 78.01 | — | |
| BIVA2021.06 | 80.06 | 78.41 | — | |
| IAF-VAE2021.06 | 80.8 | 79.1 | — | |
| BFNBound type=Upper bound, Data augmentation=false2023.08 | — | — | 0.7787 | |
| CR-NVAEResolution=28 x 282021.05 | — | 76.93 | — | |
| CR-NVAEBound type=Upper bound, Data augmentation=true2023.08 | — | — | 0.7693 | |
| DVAE++2021.06 | — | 78.49 | — | |
| Locally Masked PixelCNNBound type=Exact value, Data augmentation=false2023.08 | — | — | 0.7758 | |
| MAE2021.06 | — | 77.98 | — | |
| NVAEResolution=28 x 282021.05 | — | 78.19 | — | |
| NVAEBound type=Upper bound, Data augmentation=false2023.08 | — | — | 0.7801 | |
| NVAE+AugResolution=28 x 282021.05 | — | 77.53 | — | |
| PixelVAE++2021.06 | — | 78 | — | |
| PixelVAE++Bound type=Exact value, Data augmentation=false2023.08 | — | — | 0.78 | |
| VampPrior2021.06 | — | 78.45 | — |