Unconditional Density Estimation on MINIBOONE (test)
8.86NLL (nats)MAF-DDSF
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MAF-DDSF2018.10 | 8.86 | — | |
| MAF-DDSF2018.10 | 8.86 | — | |
| RQ-NSF (AR)Training Approach=Maximum Likelihood2020.07 | 9.22 | — | |
| GFTraining Approach=Maximum Likelihood2020.07 | 10.32 | — | |
| FFJORD2018.10 | 10.43 | — | |
| FFJORD2018.10 | 10.43 | — | |
| FFJORDTraining Approach=Maximum Likelihood2020.07 | 10.43 | — | |
| TAN2018.10 | 11.01 | — | |
| TAN2018.10 | 11.01 | — | |
| Glow2018.10 | 11.35 | — | |
| Glow2018.10 | 11.35 | — | |
| GlowTraining Approach=Maximum Likelihood2020.07 | 11.35 | — | |
| MAFTraining Approach=Maximum Likelihood2020.07 | 11.68 | — | |
| MAF2018.10 | 11.75 | — | |
| MAF2018.10 | 11.75 | — | |
| Real NVP2018.10 | 13.55 | — | |
| Real NVP2018.10 | 13.55 | — | |
| Real NVPTraining Approach=Maximum Likelihood2020.07 | 13.55 | — | |
| GISTraining Approach=Iterative2020.07 | 14.26 | — | |
| MADE2018.10 | 15.59 | — | |
| MADE2018.10 | 15.59 | — | |
| RBIGTraining Approach=Iterative2020.07 | 25.41 | — | |
| B-NAFstacked flows=52019.04 | — | -8.95 | |
| FFJORD2019.04 | — | -10.43 | |
| Gaussian2017.05 | — | -37.24 | |
| Glow2019.04 | — | -11.35 | |
| MADE2017.05 | — | -15.59 | |
| MADE MoGOutput distribution=Mixture of Gaussians2017.05 | — | -12.27 | |
| MADE MoG2019.04 | — | -12.27 | |
| MAFNumber of layers=52017.05 | — | -11.75 | |
| MAFNumber of layers=102017.05 | — | -12.24 | |
| MAF MoGNumber of layers=5, Output distribution=Mixture of Gaussians2017.05 | — | -11.68 | |
| MAF-affine2019.04 | — | -12.24 | |
| MAF-affine MoG2019.04 | — | -11.68 | |
| NAF-DDSF2019.04 | — | -8.86 | |
| Real NVPNumber of layers=52017.05 | — | -13.55 | |
| Real NVPNumber of layers=102017.05 | — | -13.84 | |
| Real NVP2019.04 | — | -13.55 | |
| TAN2019.04 | — | -11.01 |