Generative Model Assessment for Robust Optimization on Qualitative Five-Point Framework
256Number of Binary VariablesNF (RealNVP)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NF (RealNVP)MIP type=MINLP, Primary failure=(P5): multistep, MINLP2026.06 | 256 | — | — | — | — | — | |
| VAE (KL)MIP type=MILP, Primary failure=(P3): KL ̸⇒ q(z) ≈ N2026.06 | 128 | — | — | — | — | — | |
| GANMIP type=MILP, Primary failure=(P3): no encoder2026.06 | 128 | — | — | — | — | — | |
| DDIMMIP type=MILP, Primary failure=(P5): multistep2026.06 | 128 | — | — | — | — | — | |
| WAAE-ROMIP type=MILP, Primary failure=Satisfies all five2026.06 | 128 | — | — | — | — | — | |
| WAAE-GMM-ROMIP type=MILP, Primary failure=GMM-guided (P3)2026.06 | 128 | — | — | — | — | — |