Regression on Kin8nm (Avg NLL Relative Percentage and Actual NLL)
-1.03Actual NLLMCMC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MCMC2025.12 | -1.03 | 767 | |
| GPExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=322026.07 | -0.68 | — | |
| SVI2025.12 | -0.59 | 388 | |
| BDKNExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | -0.48 | — | |
| TAGI2025.12 | -0.45 | 128 | |
| Laplace2025.12 | -0.43 | 552 | |
| KBNN2025.12 | -0.41 | 821 | |
| BLLExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | -0.34 | — | |
| VBLLExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | -0.21 | — | |
| Dropout2025.12 | -0.08 | 11,821 | |
| DEExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | -0.02 | — | |
| MFVIExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | 0.01 | — | |
| LDBLLExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | 1.29 | — | |
| HABNN2025.12 | 2.87 | 22 | |
| PBP2025.12 | 275.95 | 260 | |
| DKLExperimental Setting=1 (high learning rate), Learning Rate=0.1, Epochs=40, Batch Size=32, Architecture=2 hidden layers (50 neurons each)2026.07 | 393.44 | — |