Regression on Kin8nm (NLL and RMSE)
0.07RMSECDropout
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CDropout2024.05 | 0.07 | -0.65 | — | |
| GPExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=322026.07 | 0.07 | — | — | |
| DEnsemble2024.05 | 0.09 | -1.2 | — | |
| DEvidential2024.05 | 0.09 | -1.24 | — | |
| MCDropout2024.05 | 0.1 | -0.95 | — | |
| DKLExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.1 | — | — | |
| BLLExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.1 | — | — | |
| WEvidential2024.05 | 0.15 | -0.44 | — | |
| BDKNExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons), Feature maps (H)=52026.07 | 0.15 | — | — | |
| NGBoost2024.05 | 0.16 | -0.49 | — | |
| VBLLExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.19 | — | — | |
| DEExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.22 | — | — | |
| MFVIExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.23 | — | — | |
| LDBLLExperimental setting=1 (high learning rate), Learning rate=0.1, Training epochs=40, Batch size=32, Architecture=2 hidden layers (50 neurons)2026.07 | 0.72 | — | — | |
| SVGPModel Variant=SVGP2024.08 | — | 400 | 5.19 | |
| SVGP+TModel Variant=SVGP+T2024.08 | — | 352 | 4.65 | |
| SVTP+MCModel Variant=SVTP+MC2024.08 | — | 225 | 3.85 | |
| SVTP+UBModel Variant=SVTP+UB2024.08 | — | 385 | 4.35 |