Generative Modeling on Omniglot (test)
-91.487Log LikelihoodACS
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ACSestimation=AIS2024.02 | -91.487 | — | — | — | — | — | — | — | — | — | |
| GWGestimation=AIS2024.02 | -94.72 | — | — | — | — | — | — | — | — | — | |
| DMALAestimation=AIS2024.02 | -99.243 | — | — | — | — | — | — | — | — | — | |
| LIDGMVAE2023.01 | -820.3 | 1 | 1.7 | 7.5 | — | — | — | — | — | — | |
| β-VAEbeta=0.22023.01 | -842.6 | 0.7 | 1.4 | 5.9 | — | — | — | — | — | — | |
| Lagging VAE2023.01 | -849.4 | 0.5 | 1 | 3.6 | — | — | — | — | — | — | |
| SA-VAE2023.01 | -853.4 | 0.1 | 0.2 | 1 | — | — | — | — | — | — | |
| VAE2023.01 | -862.1 | 0.02 | 0 | 0.1 | — | — | — | — | — | — | |
| ACSSampling method=Automatic Cyclical Sampler2024.02 | — | — | — | — | — | — | — | -220.71 | — | — | |
| Attentive VAEL=152024.12 | — | — | — | — | 89.5 | — | — | — | — | — | |
| BIVAL=62024.12 | — | — | — | — | 91.34 | — | — | — | — | — | |
| DMALASampling method=Discrete Metropolis-Adjusted Langevin Algorithm2024.02 | — | — | — | — | — | — | — | -222.61 | — | — | |
| DVP-VAEL=8, Averaged over 4 random seeds=true2024.12 | — | — | — | — | 89.07 | — | — | — | — | — | |
| FFJORDArchitecture=VAE2024.03 | — | — | — | — | 64.3 | 20.5 | 0.99 | — | — | — | |
| GBSampling method=Block-Gibbs2024.02 | — | — | — | — | — | — | — | -161.73 | — | — | |
| GWGSampling method=Gibbs-With-Gradients2024.02 | — | — | — | — | — | — | — | -276.81 | — | — | |
| IA-VAENumber of random seeds=102026.04 | — | — | — | — | — | — | — | — | -89.73 | 0.01 | |
| LVAEL=52024.12 | — | — | — | — | 102.11 | — | — | — | — | — | |
| OU-VAEL=52024.12 | — | — | — | — | 96.08 | — | — | — | — | — | |
| PaddingFlowArchitecture=VAE, Base Method=FFJORD, Noise=Padding-dimensional2024.03 | — | — | — | — | 64 | 20.3 | 0.988 | — | — | — | |
| VAENumber of random seeds=102026.04 | — | — | — | — | — | — | — | — | -90.3 | — | |
| VampPriorL=22024.12 | — | — | — | — | 89.76 | — | — | — | — | — |