Protein Property Prediction on multi-task experiment 700 training points (test)
0.911Spearman CorrelationCLOCK-GP+CNN-ZS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLOCK-GP+CNN-ZSTraining points=700, Kernel=non-linear2026.06 | 0.911 | 0.249 | 0.927 | 0.358 | |
| CLOCK-GPTraining points=700, Kernel=non-linear2026.06 | 0.907 | 0.254 | 0.924 | 0.367 | |
| CLOCK-GP-LIN+CNN-ZSTraining points=700, Kernel=linear2026.06 | 0.899 | 0.269 | 0.916 | 0.387 | |
| CLOCK-GP-LINTraining points=700, Kernel=linear2026.06 | 0.89 | 0.279 | 0.904 | 0.412 | |
| CNN-EMB-OHTraining points=700, Features=CNN neural features + one-hot2026.06 | 0.884 | 0.3 | 0.904 | 0.414 | |
| Global CLOCK-GPTraining points=700, Kernel=non-linear2026.06 | 0.877 | 0.296 | 0.893 | 0.436 | |
| CNN-ZS-OHTraining points=700, Features=CNN zero-shot + one-hot2026.06 | 0.863 | 0.326 | 0.887 | 0.449 | |
| Global CLOCK-GP-LINTraining points=700, Kernel=linear2026.06 | 0.859 | 0.322 | 0.874 | 0.476 | |
| LOCK-GPTraining points=700, Kernel=non-linear2026.06 | 0.848 | 0.332 | 0.862 | 0.496 | |
| LOCK-GP-LINTraining points=700, Kernel=linear2026.06 | 0.833 | 0.35 | 0.849 | 0.518 | |
| Ridge-EMB-OHTraining points=700, Features=Mean-pooled Chroma features + one-hot2026.06 | 0.829 | 0.375 | 0.849 | 0.517 | |
| Ridge-OHTraining points=700, Features=One-hot sequences2026.06 | 0.748 | 0.446 | 0.754 | 0.638 | |
| Ridge-EMBTraining points=700, Features=Mean-pooled Chroma features2026.06 | 0.746 | 0.478 | 0.773 | 0.626 | |
| CNN-ZSTraining points=700, Features=CNN zero-shot predictions2026.06 | 0.72 | 1.688 | 0.75 | 1.82 |