mlsanityData to 14 Oct 2026

Adversarial robustness of amortized Bayesian inference

arXiv 2305.14984 ↗First posted 24 May 2023cs.LG stat.MLcited by 17

Not on the map. Why?

Abstract (arXiv)

Bayesian inference usually requires running potentially costly inference procedures separately for every new observation. In contrast, the idea of amortized Bayesian inference is to initially invest computational cost in training an inference network on simulated data, which can subsequently be used to rapidly perform inference (i.e., to return estimates of posterior distributions) for new observations. This approach has been applied to many real-world models in the sciences and engineering, but it is unclear how robust the approach is to adversarial perturbations in the observed data. Here, we study the adversarial robustness of amortized Bayesian inference, focusing on simulation-based estimation of multi-dimensional posterior distributions. We show that almost unrecognizable, targeted perturbations of the observations can lead to drastic changes in the predicted posterior and highly unrealistic posterior predictive samples, across several benchmark tasks and a real-world example from neuroscience. We propose a computationally efficient regularization scheme based on penalizing the Fisher information of the conditional density estimator, and show how it improves the adversarial robustness of amortized Bayesian inference.

Key facts. Cited by 17 papers in the map (data through 14 Oct 2026). Builds on 47 of its 80 references that are in the map. First posted on arXiv 24 May 2023.

Builds on

47 of 80 references.

References · 47

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  4. 2013Intriguing properties of neural networkscited by 5,155
  5. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,517
  6. 2018Robustness May Be at Odds with Accuracycited by 880
  7. 2017Masked Autoregressive Flow for Density Estimationcited by 684
  8. 2020RobustBench: a standardized adversarial robustness benchmarkcited by 661
  9. 2019The frontier of simulation-based inferencecited by 502
  10. 2018Pyro: Deep Universal Probabilistic Programmingcited by 380
  11. 2018Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flowscited by 221
  12. 2019Automatic Posterior Transformation for Likelihood-Free Inferencecited by 219
35 more
  1. 2017Flexible statistical inference for mechanistic models of neural dynamicscited by 163
  2. 2021Benchmarking Simulation-Based Inferencecited by 162
  3. 2020BayesFlow: Learning complex stochastic models with invertible neural networkscited by 144
  4. 2014Inconsistency of Bayesian Inference for Misspecified Linear Models, and a Proposal for Repairing Itcited by 95
  5. 2019BayesSim: adaptive domain randomization via probabilistic inference for robotics simulatorscited by 85
  6. 2021Real-time gravitational-wave science with neural posterior estimationcited by 78
  7. 2015Distributional Smoothing with Virtual Adversarial Trainingcited by 61
  8. 2021Robust Generalised Bayesian Inference for Intractable Likelihoodscited by 53
  9. 2021Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networkscited by 49
  10. 2019Scaleable input gradient regularization for adversarial robustnesscited by 47
  11. 2022Robust Neural Posterior Estimation and Statistical Model Criticismcited by 46
  12. 2022Investigating the Impact of Model Misspecification in Neural Simulation-based Inferencecited by 45
  13. 2022Robust Bayesian Inference for Simulator-based Models via the MMD Posterior Bootstrapcited by 41
  14. 2022Truncated proposals for scalable and hassle-free simulation-based inferencecited by 41
  15. 2020Invertible Generative Modeling using Linear Rational Splinescited by 40
  16. 2020The Curious Case of Adversarially Robust Models: More Data Can Help, Double Descend, or Hurt Generalizationcited by 39
  17. 2018A Likelihood-Free Inference Framework for Population Genetic Data using Exchangeable Neural Networkscited by 31
  18. 2018Amortized Inference Regularizationcited by 30
  19. 2021A Review of Adversarial Attack and Defense for Classification Methodscited by 29
  20. 2021A Trust Crisis In Simulation-Based Inference? Your Posterior Approximations Can Be Unfaithfulcited by 28
  21. 2022Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inferencecited by 28
  22. 2019On the well-posedness of Bayesian inverse problemscited by 20
  23. 2018Adversarial Attacks on Variational Autoencoderscited by 19
  24. 2018The Adversarial Attack and Detection under the Fisher Information Metriccited by 18
  25. 2021On the Robustness to Misspecification of α-Posteriors and Their Variational Approximationscited by 15
  26. 2021Certifiably Robust Variational Autoencoderscited by 14
  27. 2020Adversarial Attacks on Probabilistic Autoregressive Forecasting Modelscited by 13
  28. 2023Conditional Generative Models are Provably Robust: Pointwise Guarantees for Bayesian Inverse Problemscited by 13
  29. 2022Robust Simulation-Based Inference in Cosmology with Bayesian Neural Networkscited by 11
  30. 2022Alleviating Adversarial Attacks on Variational Autoencoders with MCMCcited by 10
  31. 2019Defending Against Adversarial Attacks by Suppressing the Largest Eigenvalue of Fisher Information Matrixcited by 9
  32. 2019Adversarial Robustness of Flow-Based Generative Modelscited by 8
  33. 2023Reliable Gradient-free and Likelihood-free Prompt Tuningcited by 8
  34. 2020The Expected Jacobian Outerproduct: Theory and Empiricscited by 4
  35. 2022Surprises in adversarially-trained linear regressioncited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 6; 2025: 7; 2026: 3.

17 papers.

Sort:

  1. 2024sbi reloaded: a toolkit for simulation-based inference workflowscited by 23
  2. 2023Sensitivity-Aware Amortized Bayesian Inferencecited by 20
  3. 2025Effortless, Simulation-Efficient Bayesian Inference using Tabular Foundation Modelscited by 13
  4. 2025Robust Amortized Bayesian Inference with Self-Consistency Losses on Unlabeled Datacited by 11
  5. 2025Does Unsupervised Domain Adaptation Improve the Robustness of Amortized Bayesian Inference? A Systematic Evaluationcited by 10
  6. 2024Sourcerer: Sample-based Maximum Entropy Source Distribution Estimationcited by 9
  7. 2024Compositional simulation-based inference for time seriescited by 9
  8. 2025Simulation-based Bayesian inference under model misspecificationcited by 9
  9. 2026BayesFlow 2: Multi-Backend Amortized Bayesian Inference in Pythoncited by 9
  10. 2024Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networks: An Extended Investigationcited by 8
  11. 2024Preconditioned Neural Posterior Estimation for Likelihood-free Inferencecited by 6
  12. 2024A Comprehensive Guide to Simulation-based Inference in Computational Biologycited by 6
5 more
  1. 2025Robust Simulation-Based Inference under Missing Data via Neural Processescited by 5
  2. 2025Causal Posterior Estimationcited by 1
  3. 2025Misspecification-robust amortised simulation-based inference using variational methodscited by 0
  4. 2026Unsupervised Continual Learning for Amortized Bayesian Inferencecited by 0
  5. 2026Robust Bayesian Decision Making under Adversarial Uncertaintycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Improving the Accuracy of Amortized Model Comparison with Self-Consistency
  2. 2024Testing and Improving the Robustness of Amortized Bayesian Inference for Cognitive Models
  3. 2024All-in-one simulation-based inference
  4. 2023Leveraging Self-Consistency for Data-Efficient Amortized Bayesian Inference
  5. 2026Amortized Factor Inference Networks for Posterior Inference
  6. 2025Improving the Accuracy of Amortized Model Comparison with Self-Consistency
  7. 2025PriorGuide: Test-Time Prior Adaptation for Simulation-Based Inference
  8. 2024Towards Improved Variational Inference for Deep Bayesian Models