Off-policy value estimation on Synthetic CRM sample splitting (50% calibration fold)
0.342(1 - γ)-scaled RMSEHybrid Isotonic–Histogram Calibration
Evaluation results
| Method | Links | |
|---|---|---|
| Hybrid Isotonic–Histogram CalibrationBase Model=Neural, Sample size (n)=100,0002025.12 | 0.342 | |
| Isotonic Iterated Bellman CalibrationBase Model=Neural, Sample size (n)=100,0002025.12 | 0.351 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Boosted, Sample size (n)=100,0002025.12 | 0.353 | |
| Isotonic Iterated Bellman CalibrationBase Model=Boosted, Sample size (n)=100,0002025.12 | 0.358 | |
| Raw (No Calibration)Base Model=Boosted, Sample size (n)=100,0002025.12 | 0.36 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Neural, Sample size (n)=50,0002025.12 | 0.374 | |
| Raw (No Calibration)Base Model=Neural, Sample size (n)=100,0002025.12 | 0.379 | |
| Isotonic Iterated Bellman CalibrationBase Model=Neural, Sample size (n)=50,0002025.12 | 0.383 | |
| Raw (No Calibration)Base Model=Neural, Sample size (n)=50,0002025.12 | 0.419 | |
| Isotonic Iterated Bellman CalibrationBase Model=Boosted, Sample size (n)=50,0002025.12 | 0.429 | |
| Raw (No Calibration)Base Model=Boosted, Sample size (n)=50,0002025.12 | 0.433 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Boosted, Sample size (n)=50,0002025.12 | 0.438 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Neural, Sample size (n)=10,0002025.12 | 0.52 | |
| Isotonic Iterated Bellman CalibrationBase Model=Neural, Sample size (n)=10,0002025.12 | 0.55 | |
| Raw (No Calibration)Base Model=Neural, Sample size (n)=10,0002025.12 | 0.582 | |
| Isotonic Iterated Bellman CalibrationBase Model=Linear, Sample size (n)=100,0002025.12 | 0.605 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Linear, Sample size (n)=100,0002025.12 | 0.606 | |
| Isotonic Iterated Bellman CalibrationBase Model=Linear, Sample size (n)=10,0002025.12 | 0.612 | |
| Isotonic Iterated Bellman CalibrationBase Model=Linear, Sample size (n)=50,0002025.12 | 0.614 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Linear, Sample size (n)=50,0002025.12 | 0.616 | |
| Raw (No Calibration)Base Model=Linear, Sample size (n)=100,0002025.12 | 0.616 | |
| Raw (No Calibration)Base Model=Linear, Sample size (n)=50,0002025.12 | 0.625 | |
| Raw (No Calibration)Base Model=Linear, Sample size (n)=10,0002025.12 | 0.64 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Linear, Sample size (n)=10,0002025.12 | 0.641 | |
| Isotonic Iterated Bellman CalibrationBase Model=Boosted, Sample size (n)=10,0002025.12 | 0.671 | |
| Raw (No Calibration)Base Model=Boosted, Sample size (n)=10,0002025.12 | 0.681 | |
| Hybrid Isotonic–Histogram CalibrationBase Model=Boosted, Sample size (n)=10,0002025.12 | 0.697 |