Generative Modeling on Omniglot dynamically binarized (test)
-107.26Log-Likelihood Bound (100-point)RODEO
Evaluation results
| Method | Links | |
|---|---|---|
| RODEONumber of layers=Three layers2022.02 | -107.26 | |
| DisARMNumber of layers=Three layers2022.02 | -107.32 | |
| Double CVNumber of layers=Three layers2022.02 | -107.48 | |
| RODEONumber of layers=Four layers2022.02 | -107.77 | |
| Double CVNumber of layers=Four layers2022.02 | -107.86 | |
| DisARMNumber of layers=Four layers2022.02 | -107.87 | |
| DisARMNumber of layers=Two layers2022.02 | -110.15 | |
| RODEONumber of layers=Two layers2022.02 | -110.32 | |
| Double CVNumber of layers=Two layers2022.02 | -110.48 |