Simulation-based inference on DDM d=3 (test)
101Gap Closed toward Oracle (%)MA-SBI
Evaluation results
| Method | Links | |
|---|---|---|
| MA-SBIz only=true, n_cal=5002026.06 | 101 | |
| Oracletype=neural2026.06 | 100 | |
| NNPEvariant=noise-aug, requires ground-truth parameter pairs=true, n_cal=5002026.06 | 99 | |
| RoPE-EMDrequires ground-truth parameter pairs=true, n_cal=5002026.06 | 80 | |
| NPE+z-concatz as NPE input=true, requires ground-truth parameter pairs=true, n_cal=5002026.06 | 24 | |
| RoPE-Sinkhorntuned epsilon=true, requires ground-truth parameter pairs=true, n_cal=5002026.06 | 2 | |
| NPEmode=baseline, n_cal=5002026.06 | 0 | |
| g(z) -> thetaregression=direct, requires ground-truth parameter pairs=true, n_cal=5002026.06 | -824 | |
| g(z) -> yregression=direct, requires ground-truth parameter pairs=true, n_cal=5002026.06 | -1,025 |