Density Estimation on POWER d=6; N=2,049,280 (test)
-1.78Avg Test NLLnMDMA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| nMDMA2024.10 | -1.78 | — | |
| LFM2024.10 | -0.715 | — | |
| BNAF2024.10 | -0.61 | — | |
| InterFlow2024.10 | -0.57 | — | |
| AdaCat2024.10 | -0.56 | — | |
| CPF2024.10 | -0.52 | — | |
| ScoreSDE2024.10 | -0.47 | — | |
| FFJORD# Param=43K, L (number of layers)=5, Training Batch size=100002022.12 | -0.46 | — | |
| FFJORD2024.10 | -0.46 | — | |
| JKO-iFlow# Param=95K, L (number of layers)=5, Training #Batches=6.08K, Training Batch size=100002022.12 | -0.4 | — | |
| JKO-iFlow2024.10 | -0.4 | — | |
| OT-Flow# Param=18K, Training #Batches=22K, Training Batch size=100002022.12 | -0.3 | — | |
| OT-Flow2024.10 | -0.3 | — | |
| Roundtrip2024.10 | 4.33 | — | |
| MADE2018.01 | — | -3.08 | |
| MADE MOG2018.01 | — | 0.4 | |
| MAFNumber of transformations=52018.01 | — | 0.14 | |
| MAFNumber of transformations=102018.01 | — | 0.24 | |
| MAF MOGNumber of transformations=52018.01 | — | 0.3 | |
| Real NVPNumber of transformations=52018.01 | — | -0.02 | |
| Real NVPNumber of transformations=102018.01 | — | 0.17 | |
| TANModel architecture=5x L+ReLU+SRNN+Re & RAM2018.01 | — | 0.6 |