mlsanityData to 14 Oct 2026

Robust Expected Information Gain for Optimal Bayesian Experimental Design Using Ambiguity Sets

arXiv 2205.09914 ↗First posted 20 May 2022stat.ML cs.AI cs.LGcited by 12

Not on the map. Why?

Abstract (arXiv)

The ranking of experiments by expected information gain (EIG) in Bayesian experimental design is sensitive to changes in the model's prior distribution, and the approximation of EIG yielded by sampling will have errors similar to the use of a perturbed prior. We define and analyze robust expected information gain (REIG), a modification of the objective in EIG maximization by minimizing an affine relaxation of EIG over an ambiguity set of distributions that are close to the original prior in KL-divergence. We show that, when combined with a sampling-based approach to estimating EIG, REIG corresponds to a `log-sum-exp' stabilization of the samples used to estimate EIG, meaning that it can be efficiently implemented in practice. Numerical tests combining REIG with variational nested Monte Carlo (VNMC), adaptive contrastive estimation (ACE) and mutual information neural estimation (MINE) suggest that in practice REIG also compensates for the variability of under-sampled estimators.

Key facts. Cited by 12 papers in the map (data through 14 Oct 2026). Builds on 7 of its 21 references that are in the map. First posted on arXiv 20 May 2022.

Builds on

7 of 21 references.

References · 7

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2018Pyro: Deep Universal Probabilistic Programmingcited by 380
  3. 2019Variational Bayesian Optimal Experimental Designcited by 116
  4. 2019A Unified Stochastic Gradient Approach to Designing Bayesian-Optimal Experimentscited by 58
  5. 2020Bayesian Experimental Design for Implicit Models by Mutual Information Neural Estimationcited by 56
  6. 2021Bayesian Distributionally Robust Optimizationcited by 16
  7. 2014Approximate Models and Robust Decisionscited by 10

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 3; 2024: 0; 2025: 5; 2026: 4.

12 papers.

  1. 2023Modern Bayesian Experimental Designcited by 125
  2. 2025Step-DAD: Semi-Amortized Policy-Based Bayesian Experimental Designcited by 12
  3. 2023Metrics for Bayesian Optimal Experiment Design under Model Misspecificationcited by 7
  4. 2023Bayesian Active Learning in the Presence of Nuisance Parameterscited by 6
  5. 2025Towards Practical Multi-label Causal Discovery in High-Dimensional Event Sequences via One-Shot Graph Aggregationcited by 4
  6. 2025Performance Comparisons of Reinforcement Learning Algorithms for Sequential Experimental Designcited by 3
  7. 2025Bayesian Experimental Design for Model Discrepancy Calibration: An Auto-Differentiable Ensemble Kalman Inversion Approachcited by 2
  8. 2026Sequential Bayesian Optimal Experimental Design in Infinite Dimensions via Policy Gradient Reinforcement Learningcited by 1
  9. 2026Maximin Robust Bayesian Experimental Designcited by 1
  10. 2025Posterior-driven Heuristic Support Adaptation in a Probabilistic Treatment of Real2Sim2Real for Vision-Driven Deformable Linear Object Manipulationcited by 0
  11. 2026Bayesian Experimental Design for Model Discrepancy Calibration: A Rivalry between Kullback--Leibler Divergence and Wasserstein Distancecited by 0
  12. 2026Robust Bayesian Decision Making under Adversarial Uncertaintycited by 0