mlsanityData to 14 Oct 2026

Misspecification-robust amortised simulation-based inference using variational methods

arXiv 2509.05724 ↗First posted 6 Sep 2025stat.ML astro-ph.GA cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Recent advances in neural density estimation have enabled powerful simulation-based inference (SBI) methods that can flexibly approximate Bayesian inference for intractable stochastic models. Although these methods have demonstrated reliable posterior estimation when the simulator accurately represents the underlying data generative process (DGP), recent work has shown that they perform poorly in the presence of model misspecification. This poses a significant issue for their use in real-world problems, due to simulators always misrepresenting the true DGP to a certain degree. In this paper, we introduce robust variational neural posterior estimation (RVNP), a method which addresses the problem of misspecification in amortised SBI by bridging the simulation-to-reality gap using variational inference and error modelling. We test RVNP on multiple benchmark tasks, including using real data from astronomy, and show that it can recover robust posterior inference in a data-driven manner without adopting hyperparameters or priors governing the misspecification influence.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 40 of its 68 references that are in the map. First posted on arXiv 6 Sep 2025.

Builds on

40 of 68 references.

References · 40

  1. 2018Learning deep representations by mutual information estimation and maximizationcited by 1,273
  2. 2017Masked Autoregressive Flow for Density Estimationcited by 684
  3. 2019Neural Spline Flowscited by 596
  4. 2019The frontier of simulation-based inferencecited by 502
  5. 2015Importance Weighted Autoencoderscited by 456
  6. 2018Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flowscited by 221
  7. 2019Automatic Posterior Transformation for Likelihood-Free Inferencecited by 219
  8. 2017Flexible statistical inference for mechanistic models of neural dynamicscited by 163
  9. 2020Flows for simultaneous manifold learning and density estimationcited by 118
  10. 2020On Contrastive Learning for Likelihood-free Inferencecited by 98
  11. 2015Approximating Likelihood Ratios with Calibrated Discriminative Classifierscited by 90
  12. 2020Neural Approximate Sufficient Statistics for Implicit Modelscited by 74
28 more
  1. 2024All-in-one simulation-based inferencecited by 74
  2. 2018Likelihood-free inference with emulator networkscited by 61
  3. 2016Fast ε-free Inference of Simulation Models with Bayesian Conditional Density Estimationcited by 52
  4. 2023Learning Robust Statistics for Simulation-based Inference under Model Misspecificationcited by 48
  5. 2022Robust Neural Posterior Estimation and Statistical Model Criticismcited by 46
  6. 2022Investigating the Impact of Model Misspecification in Neural Simulation-based Inferencecited by 45
  7. 2022Variational methods for simulation-based inferencecited by 43
  8. 2018Doubly Reparameterized Gradient Estimators for Monte Carlo Objectivescited by 41
  9. 2022Robust Bayesian Inference for Simulator-based Models via the MMD Posterior Bootstrapcited by 41
  10. 2017Reinterpreting Importance-Weighted Autoencoderscited by 37
  11. 2021Sequential Neural Posterior and Likelihood Approximationcited by 30
  12. 2023Bayesian Synthetic Likelihoodcited by 29
  13. 2023Amortized Variational Inference: When and Why?cited by 29
  14. 2022Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inferencecited by 28
  15. 2024Addressing Misspecification in Simulation-based Inference through Data-driven Calibrationcited by 27
  16. 2014High-Dimensional Density Ratio Estimation with Extensions to Approximate Likelihood Computationcited by 19
  17. 2023Adversarial robustness of amortized Bayesian inferencecited by 17
  18. 2019Robust Approximate Bayesian Inference with Synthetic Likelihoodcited by 14
  19. 2021A neural simulation-based inference approach for characterizing the Galactic Center γ-ray excesscited by 12
  20. 2022Robust Simulation-Based Inference in Cosmology with Bayesian Neural Networkscited by 11
  21. 2025Robust Amortized Bayesian Inference with Self-Consistency Losses on Unlabeled Datacited by 11
  22. 2025Does Unsupervised Domain Adaptation Improve the Robustness of Amortized Bayesian Inference? A Systematic Evaluationcited by 10
  23. 2022Approximate Bayesian Computation with Domain Expert in the Loopcited by 9
  24. 2025Simulation-based Bayesian inference under model misspecificationcited by 9
  25. 2024Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networks: An Extended Investigationcited by 8
  26. 2024Domain-Adaptive Neural Posterior Estimation for Strong Gravitational Lens Analysiscited by 7
  27. 2021Synthetic likelihood in misspecified modelscited by 5
  28. 2024Variational Autoencoders for Efficient Simulation-Based Inferencecited by 2

Built on by

Not cited yet.