mlsanityData to 7 Oct 2026

Conservative neural posterior estimation via distributionally robust training

arXiv 2605.28516 ↗First posted 27 May 2026stat.ML cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Simulation-based inference with neural posterior estimation (NPE) often yields overconfident and unreliable posteriors under limited simulation budgets. To address this, we propose DRO-NPE, a distributionally robust approach that replaces the standard NPE objective with a worst-case loss over a Wasserstein ambiguity set. We introduce KL-based metrics for miscoverage and miscalibration, and use these to show that the DRO-NPE objective controls overfitting and reduces posterior overconfidence. Our method is tractable, parallelisable, and readily integrates with standard normalising flows. Across benchmark SBI tasks, DRO-NPE consistently improves coverage and calibration, while narrowing the gap between empirical and population NPE loss, leading to more reliable inference in low-simulation regimes.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 49 of its 79 references that are in the map. First posted on arXiv 27 May 2026.

Builds on

49 of 79 references.

References · 49

  1. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,539
  2. 2019Normalizing Flows for Probabilistic Modeling and Inferencecited by 1,315
  3. 2017Masked Autoregressive Flow for Density Estimationcited by 680
  4. 2019The frontier of simulation-based inferencecited by 490
  5. 2015Data-driven Distributionally Robust Optimization Using the Wasserstein Metric: Performance Guarantees and Tractable Reformulationscited by 285
  6. 2021User-friendly introduction to PAC-Bayes boundscited by 234
  7. 2018Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flowscited by 217
  8. 2019Automatic Posterior Transformation for Likelihood-Free Inferencecited by 216
  9. 2016Distributionally Robust Stochastic Optimization with Wasserstein Distancecited by 182
  10. 2017Flexible statistical inference for mechanistic models of neural dynamicscited by 162
  11. 2021Benchmarking Simulation-Based Inferencecited by 160
  12. 2020BayesFlow: Learning complex stochastic models with invertible neural networkscited by 139
37 more
  1. 2018Validating Bayesian Inference Algorithms with Simulation-Based Calibrationcited by 118
  2. 2019Distributionally Robust Optimization: A Reviewcited by 75
  3. 2020Finite-Sample Guarantees for Wasserstein Distributionally Robust Optimization: Breaking the Curse of Dimensionalitycited by 49
  4. 2022Robust Neural Posterior Estimation and Statistical Model Criticismcited by 45
  5. 2023Sampling-Based Accuracy Testing of Posterior Estimators for General Inferencecited by 45
  6. 2020PAC-Bayes unleashed: generalisation bounds with unbounded lossescited by 43
  7. 2025Simulation-Based Inference: A Practical Guidecited by 30
  8. 2021A Trust Crisis In Simulation-Based Inference? Your Posterior Approximations Can Be Unfaithfulcited by 27
  9. 2022Towards Reliable Simulation-Based Inference with Balanced Neural Ratio Estimationcited by 27
  10. 2021Diagnostics for Conditional Density Models and Bayesian Inference Algorithmscited by 26
  11. 2021Graphical Test for Discrete Uniformity and its Applications in Goodness of Fit Evaluation and Multiple Sample Comparisoncited by 26
  12. 2023A Large-Scale Study of Probabilistic Calibration in Neural Network Regressioncited by 26
  13. 2021Benchmarking Invertible Architectures on Inverse Problemscited by 25
  14. 2024Distributionally Robust Optimization and Robust Statisticscited by 24
  15. 2023Consistency Models for Scalable and Fast Simulation-Based Inferencecited by 22
  16. 2022Simulation-Based Calibration Checking for Bayesian Computation: The Choice of Test Quantities Shapes Sensitivitycited by 19
  17. 2022Distributionally Robust Bayesian Optimization with varphi-divergencescited by 16
  18. 2022Simulator-Based Inference with Waldo: Confidence Regions by Leveraging Prediction Algorithms and Posterior Estimators for Inverse Problemscited by 14
  19. 2023Balancing Simulation-based Inference for Conservative Posteriorscited by 10
  20. 2023Calibrating Neural Simulation-Based Inference with Differentiable Coverage Probabilitycited by 10
  21. 2024On Regularization via Early Stopping for Least Squares Regressioncited by 10
  22. 2021Generalization Bounds with Minimal Dependency on Hypothesis Class via Distributionally Robust Optimizationcited by 9
  23. 2022Towards Instance-Wise Calibration: Local Amortized Diagnostics and Reshaping of Conditional Densities (LADaR)cited by 9
  24. 2025Multifidelity Simulation-based Inference for Computationally Expensive Simulatorscited by 9
  25. 2020Quantile Regularization: Towards Implicit Calibration of Regression Modelscited by 8
  26. 2022Validation Diagnostics for SBI algorithms based on Normalizing Flowscited by 8
  27. 2025Bifidelity Parameter Estimation Using Conditional Diffusion Modelscited by 7
  28. 2026BayesFlow 2: Multi-Backend Amortized Bayesian Inference in Pythoncited by 7
  29. 2022The CAMELS project: public data releasecited by 6
  30. 2024Cost-aware simulation-based inferencecited by 6
  31. 2024Low-Budget Simulation-Based Inference with Bayesian Neural Networkscited by 5
  32. 2025Multilevel neural simulation-based inferencecited by 5
  33. 2025CP4SBI: Local Conformal Calibration of Credible Sets in Simulation-Based Inferencecited by 5
  34. 2023Learning Against Distributional Uncertainty: On the Trade-off Between Robustness and Specificitycited by 4
  35. 2025Transfer learning for multifidelity simulation-based inference in cosmologycited by 4
  36. 2023Simulation-based stackingcited by 3
  37. 2026Amortised and provably-robust simulation-based inferencecited by 3

Built on by

Not cited yet.