Unconditional Density Estimation on HEPMASS (test)
14.01NLL (nats)RQ-NSF (AR)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RQ-NSF (AR)Training Approach=Maximum Likelihood2020.07 | 14.01 | — | |
| FFJORD2018.10 | 14.92 | — | |
| FFJORD2018.10 | 14.92 | — | |
| FFJORDTraining Approach=Maximum Likelihood2020.07 | 14.92 | — | |
| MAF-DDSF2018.10 | 15.09 | — | |
| MAF-DDSF2018.10 | 15.09 | — | |
| TAN2018.10 | 15.12 | — | |
| TAN2018.10 | 15.12 | — | |
| MAFTraining Approach=Maximum Likelihood2020.07 | 17.39 | — | |
| GFTraining Approach=Maximum Likelihood2020.07 | 17.59 | — | |
| MAF2018.10 | 17.7 | — | |
| MAF2018.10 | 17.7 | — | |
| Real NVP2018.10 | 18.71 | — | |
| Real NVP2018.10 | 18.71 | — | |
| Real NVPTraining Approach=Maximum Likelihood2020.07 | 18.71 | — | |
| Glow2018.10 | 18.92 | — | |
| Glow2018.10 | 18.92 | — | |
| GlowTraining Approach=Maximum Likelihood2020.07 | 18.92 | — | |
| GISTraining Approach=Iterative2020.07 | 19 | — | |
| MADE2018.10 | 20.98 | — | |
| MADE2018.10 | 20.98 | — | |
| RBIGTraining Approach=Iterative2020.07 | 24.59 | — | |
| DDE (Ours)Normalization Strategy=Sampling, sigma_n=0.022020.01 | — | -11.3 | |
| De Cao et al. (2019)Normalization Strategy=Sampling2020.01 | — | -14.71 | |
| Dinh et al. (2017)Normalization Strategy=Analytical2020.01 | — | -18.71 | |
| Gaussian2017.05 | — | -27.93 | |
| Germain et al. (2015)Normalization Strategy=Analytical, Mixture of Gaussians=true2020.01 | — | -15.15 | |
| Grathwohl et al. (2019)Normalization Strategy=Analytical2020.01 | — | -14.92 | |
| Huang et al. (2018)Normalization Strategy=Analytical2020.01 | — | -15.09 | |
| Kingma and Dhariwal (2018)Normalization Strategy=Analytical2020.01 | — | -18.92 | |
| MADE2017.05 | — | -20.98 | |
| MADE MoGOutput distribution=Mixture of Gaussians2017.05 | — | -15.15 | |
| MAFNumber of layers=52017.05 | — | -17.7 | |
| MAFNumber of layers=102017.05 | — | -17.73 | |
| MAF MoGNumber of layers=5, Output distribution=Mixture of Gaussians2017.05 | — | -17.39 | |
| Oliva et al. (2018)Normalization Strategy=Analytical2020.01 | — | -13.78 | |
| Papamakarios et al. (2017)Normalization Strategy=Analytical2020.01 | — | -17.73 | |
| Papamakarios et al. (2017)Normalization Strategy=Analytical, Mixture of Gaussians=true2020.01 | — | -17.39 | |
| Real NVPNumber of layers=52017.05 | — | -19.62 | |
| Real NVPNumber of layers=102017.05 | — | -18.71 |