Optimal Policy Recovery (Empirical Coverage) on RiverSwim T=50 nominal 95% coverage
95.7Q* Recovery (s=1, a=0)Model-based bootstrap with percentile CI
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Model-based bootstrap with percentile CIn=5002026.05 | 95.7 | 95.3 | 97.1 | 95.7 | 95.6 | 95.3 | 95.8 | 96.4 | 97 | |
| Model-based bootstrap with percentile CIn=10002026.05 | 93.8 | 94.5 | 96 | 93.8 | 93.3 | 94.5 | 94.5 | 94.7 | 95 | |
| Episodic bootstrap with percentile CIn=10002026.05 | 90.5 | 91.1 | 92.3 | 90.5 | 90.4 | 91.1 | 91.1 | 91.2 | 91.1 | |
| Episodic bootstrap with percentile CIn=5002026.05 | 88.6 | 89.3 | 90.6 | 88.6 | 89 | 89.3 | 88.2 | 88 | 89.9 | |
| Model-based bootstrap with percentile CIn=1002026.05 | 66.7 | 67 | 50.1 | 66.7 | 66.8 | 67 | 66.8 | 66.9 | 55.5 | |
| Plug-in (CLT) CIn=10002026.05 | 49.7 | 88.5 | 92.7 | 49.7 | 84.6 | 88.5 | 91.4 | 92.4 | 92.5 | |
| Plug-in (CLT) CIn=5002026.05 | 48.9 | 86.4 | 89.5 | 48.9 | 77.2 | 86.4 | 88 | 88.2 | 88 | |
| Model-based bootstrap with pivot CIn=10002026.05 | 47.1 | 80.1 | 88.7 | 47.1 | 67.6 | 78.1 | 85.6 | 89.3 | 90.1 | |
| Model-based bootstrap with pivot CIn=5002026.05 | 46.8 | 73.5 | 83.1 | 46.8 | 61.3 | 70.7 | 78.4 | 82.7 | 83.4 | |
| Episodic bootstrap with pivot CIn=10002026.05 | 45.3 | 78.3 | 84.5 | 45.3 | 65.7 | 76.4 | 82.6 | 84.9 | 86.1 | |
| Episodic bootstrap with pivot CIn=5002026.05 | 44.4 | 69.8 | 77.6 | 44.4 | 59 | 67.6 | 74.9 | 78.3 | 77.6 | |
| Episodic bootstrap with percentile CIn=1002026.05 | 43.4 | 43.4 | 22.5 | 43.4 | 43.2 | 43.8 | 43 | 42.8 | 29.3 | |
| Model-based bootstrap with percentile CIn=502026.05 | 43.3 | 43.3 | 28.9 | 43.3 | 43.3 | 43.3 | 43 | 42.9 | 32.5 | |
| Plug-in (CLT) CIn=1002026.05 | 37.3 | 52.1 | 51.4 | 37.3 | 46.5 | 52.1 | 52.6 | 51.5 | 50 | |
| Model-based bootstrap with pivot CIn=1002026.05 | 36.1 | 55.5 | 50.3 | 36.1 | 49.8 | 55.4 | 50.3 | 50.9 | 46.6 | |
| Plug-in (CLT) CIn=502026.05 | 25.1 | 32.1 | 31.3 | 25.1 | 30.6 | 31.8 | 32.3 | 32.1 | 30.6 | |
| Model-based bootstrap with pivot CIn=502026.05 | 24.6 | 72 | 30.6 | 24.6 | 60.9 | 74.2 | 31.6 | 30.9 | 28.6 | |
| Episodic bootstrap with pivot CIn=1002026.05 | 15.9 | 25.1 | 36.9 | 15.9 | 22.5 | 24.7 | 21.3 | 22.5 | 29 | |
| Episodic bootstrap with pivot CIn=502026.05 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
| Episodic bootstrap with percentile CIn=502026.05 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |