Constrained Optimization (Linear Objective) on Two Moons 1,500 problems (test)
100Feasibility (%)FAB: 2 Decoder Nets
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FAB: 2 Decoder NetsType=ML-Based2026.04 | 100 | 0.55 | 0.94 | |
| FAB: 3 Decoder NetsType=ML-Based2026.04 | 100 | 0.62 | 0.9 | |
| FSNetType=ML-Based2026.04 | 98.3 | 3.4 | 0.96 | |
| FABType=ML-Based2026.04 | 83.5 | 0.43 | 0.73 | |
| Penalty NNType=ML-Based2026.04 | 80 | 0.14 | 0.85 | |
| Interior PointType=Classical2026.04 | 77.3 | 46.37 | 1.03 | |
| Projected GradientType=Classical2026.04 | 76.3 | 48.23 | 0.9 | |
| Homeomorphic ProjectionType=ML-Based2026.04 | 76.1 | 240 | 1.67 | |
| FAB: Phase 1 OnlyType=ML-Based2026.04 | 69.9 | 1.07 | 0.72 | |
| FAB: No DξType=ML-Based2026.04 | 54.4 | 0.67 | 0.7 | |
| FAB: No Hinge LossType=ML-Based2026.04 | 18.8 | 0.76 | 0.85 | |
| Augmented LagrangianType=Classical2026.04 | 10.9 | 56.98 | 1.56 | |
| Penalty MethodType=Classical2026.04 | 10.4 | 61.58 | 1.52 |