Generative Modeling on SVHN (test)
2.15Bits Per DimensionEBM-FCE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EBM-FCETraining objective=Flow Contrastive Estimation, Normalizing constant=Estimated2019.12 | 2.15 | — | — | — | |
| Glow-MLETraining objective=Maximum Likelihood Estimation2019.12 | 2.17 | — | — | — | |
| Glow-FCETraining objective=Flow Contrastive Estimation2019.12 | 2.25 | — | — | — | |
| C-NCSN++2022.09 | — | 5 | 0 | 0 | |
| QC-NCSN++2022.09 | — | 2.01 | 16,900,000 | 0.0005 | |
| U-NCSN++2022.09 | — | 2.15 | 30,600,000 | 0.0007 |