Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Robust Statistics for Simulation-based Inference under Model Misspecification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add MMD regularizer on learned statistics (NPE -> NPE-RS) | NPE · Ricker contamination 20% | RMSE: 11.2 -> 4.66 | Helped |
| add MMD regularizer (NPE -> NPE-RS) | NPE · OUP contamination 20% | RMSE: 2.59 -> 0.63 | Helped |
| add MMD regularizer (ABC -> ABC-RS) | ABC · Ricker contamination 20% | RMSE: 9.79 -> 2.99 | Helped |
| add MMD regularizer (ABC -> ABC-RS) | ABC · OUP contamination 20% | RMSE: 5.16 -> 0.88 | Helped |
| add MMD regularizer (NPE -> NPE-RS) | NPE · Ricker well-specified | RMSE: 2.16 -> 2.18 | No clear effect |
| add MMD regularizer (ABC -> ABC-RS) | ABC · Ricker well-specified | RMSE: 1.46 -> 1.20 | No clear effect |
| NPE-RS vs robust NPE (RNPE) | NPE-RS · Ricker contamination 20% | RMSE: 7.14 -> 4.66 | Helped |
| NPE-RS vs RNPE | NPE-RS · Turin real data | MMD: 0.09 -> 0.03 | Helped |
| use Euclidean distance instead of MMD in margin regularizer (ablation) | NPE-RS · Ricker well-specified | posterior quality | Hurt |
| increase lambda to 1000 (ablation) | NPE-RS · Ricker contamination 20% | posterior vs prior MMD | Hurt |
| regularize with lambda 20-100 under adversarial training (NPE -> NPE-RS) | NPE · Gaussian linear model | MMD: 0.26 -> 0.18 | Helped |
| add MMD regularizer under prior misspecification | NPE · Ricker lognormal prior | posterior coverage of true theta | Helped |
Closest papers by meaning
- Addressing Misspecification in Simulation-based Inference through Data-driven Calibration (ICML 2025 · similarity 0.91)
- Robust Simulation-Based Inference under Missing Data via Neural Processes (ICLR 2025 · similarity 0.88)
- Consistency Models for Scalable and Fast Simulation-Based Inference (NeurIPS 2024 · similarity 0.87)
- Multilevel neural simulation-based inference (NeurIPS 2025 · similarity 0.87)
- Compositional simulation-based inference for time series (ICLR 2025 · similarity 0.87)
- Effortless, Simulation-Efficient Bayesian Inference using Tabular Foundation Models (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.