Monotonicity Regression on Synthetic 1D Monotone evaluation grid pi_eval
0.078Delta MAEBaseline
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| BaselinePrimary Network=Width-16 four-layer softplus MLP2026.06 | 0.078 | 0.24 | 0.14 | 1.2 | 0 | 22.8 | |
| CMNNPrimary Network=Width-16 four-layer softplus MLP2026.06 | 0.085 | 0 | 0 | 0 | 100 | 27.6 | |
| LagrangianPrimary Network=Width-16 four-layer softplus MLP2026.06 | 0.091 | 0.02 | 0 | 0.0017 | 2 | 29.4 | |
| PenaltyPrimary Network=Width-16 four-layer softplus MLP2026.06 | 0.092 | 0.006 | 0 | 0.0003 | 2 | 32.4 | |
| Aug. Lag.Primary Network=Width-16 four-layer softplus MLP2026.06 | 0.093 | 0.0048 | 0 | 0.0002 | 73 | 32.9 | |
| N. Slack Var.Primary Network=Width-16 four-layer softplus MLP, Slack Architecture=SIREN (width-16, 3 layers)2026.06 | 0.096 | 0 | 0 | 0 | 100 | 44.7 |