Parameter Estimation on Multiscale Lorenz-96 (test)
6.3Mean AP Error (F)h_theta w/ EMBED & EMULATE
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| h_theta w/ EMBED & EMULATETraining size (n)=10002022.11 | 6.3 | 1.86 | 2.07 | 1.31 | 9.34 | 3.71 | 5.51 | 1.74 | |
| EnKI w/ EMBED & EMULATETraining size (n)=10002022.11 | 6.59 | 2.14 | 2.36 | 1.54 | 9.38 | 4.02 | 5.54 | 2.35 | |
| Supervised RegressionTraining size (n)=10002022.11 | 7.57 | 1.78 | 3.08 | 1.54 | 11.46 | 3.29 | 6.64 | 2.19 | |
| EnKI w/ EMBED & EMULATETraining size (n)=5002022.11 | 10.94 | 4.07 | 3.74 | 2.26 | 16.09 | 6.41 | 8.84 | 2.93 | |
| h_theta w/ EMBED & EMULATETraining size (n)=5002022.11 | 11.19 | 3.65 | 3.18 | 1.6 | 15.52 | 6.24 | 8.57 | 2.17 | |
| NPE-CTraining size (n)=10002022.11 | 11.29 | 3.54 | 5.62 | 2.29 | 16.32 | 6.53 | 6.81 | 2.38 | |
| Supervised RegressionTraining size (n)=5002022.11 | 11.97 | 3.1 | 4.07 | 2.24 | 17.04 | 5.88 | 9.07 | 2.74 | |
| EnKI w/o LearningTraining size (n)=02022.11 | 15.48 | 3.77 | 0.86 | 0.2 | 40.45 | 13.39 | 4.6 | 0.6 | |
| NPE-CTraining size (n)=5002022.11 | 15.51 | 4.52 | 5.94 | 2.59 | 23.54 | 8.16 | 9.42 | 3.7 | |
| SNL+Training size (n)=10002022.11 | 33.29 | 19.68 | 49.05 | 27.75 | 48.17 | 23.9 | 25.4 | 16.03 | |
| SNL+Training size (n)=5002022.11 | 36.88 | 19.93 | 48.71 | 30.69 | 57.45 | 28.59 | 23.45 | 17.62 |