Table 1:
Analysis Results for the T2DM Dataset. Estimated averages and standard deviations (in parenthesis) of the value function are reported using 5-fold cross-validation with 50 replications. OWL-1 and OWL-2 represent two extensions of OWL (one-versus-rest and one-versus-one), MOML and MOML-l1 represent the outcome weighted margin-based learning with l2 and l1 penalties respectively. The observed average reward for the cleaned dataset is 2.246.
| Training | Validation | |
|---|---|---|
| OWL-1-Linear | 2.712 (0.329) | 2.371 (0.483) |
| OWL-2-Linear | 2.487 (0.233) | 2.221 (0.561) |
| OWL-1-Gaussian | 4.118 (0.401) | 3.285 (0.490) |
| OWL-2-Gaussian | 4.003 (0.374) | 3.221 (0.468) |
| MOML-Linear | 2.610 (0.130) | 2.440 (0.320) |
| MOML-l1-Linear | 2.813 (0.138) | 2.533 (0.182) |
| MOML-Gaussian | 4.105 (0.221) | 3.612 (0.328) |