Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Variational Bayesian Optimal Experimental Design
Adam Foster, Martin Jankowiak, Elias Bingham, Paul Horsfall, Yee Whye Teh, Thomas Rainforth, Noah Goodman
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use the variational EIG estimators instead of NMC | EIG estimation across A/B test, Preference, Mixed effects, and Extrapolation · Four EIG-estimation benchmarks | EIG estimation accuracy | Improved |
| Use VNMC instead of NMC | EIG estimation for A/B test and Preference · A/B test and Preference | Bias^2: 4.70\times 10^0 -> 3.44\times 10^{-3}; 7.60\times 10^{-2} -> 4.17\times 10^{-3} | Improved |
| Use IML instead of DV | EIG estimation for Mixed effects · Mixed effects | Bias^2: 9.10 \times 10^{-3} -> 3.06\times 10^{-3} | Improved |
| Use IML instead of DV | EIG estimation for Extrapolation · Extrapolation | Var: 4.11\times 10^{-5} -> 1.84\times 10^{-5} | Improved |
| Use Laplace instead of variational posterior | EIG estimation for A/B test · A/B test | Bias^2: 1.33\times 10^{-2} -> 1.92\times 10^{-4} | Improved |
| Use variational marginal instead of variational posterior | EIG estimation for Preference · Preference | Bias^2: 4.26\times 10^{-2} -> 1.10\times 10^{-3} | Improved |
| Train the VNMC proposal before the NMC refinement | VNMC EIG estimation · A/B test example | EIG estimation accuracy | Improved |
| Increase the VNMC refinement sample counts N and M after proposal training | VNMC EIG estimation · A/B test example | EIG estimation accuracy | Improved |
| Allocate part of the compute budget to variational training rather than using the prior as the NMC proposal | Sequential CES experiment · CES parameter-recovery experiment | Posterior entropy | Improved |
| Use BOED with IML instead of random design | Sequential adaptive psychology experiment · Mechanical Turk face-feature experiment | Posterior entropy | Improved |
Closest papers by meaning
- Bayesian Experimental Design Via Contrastive Diffusions (ICLR 2025 · similarity 0.92)
- On Estimating the Gradient of the Expected Information Gain in Bayesian Experimental Design (AAAI 2024 · similarity 0.90)
- Maximin Robust Bayesian Experimental Design (NeurIPS 2026 · similarity 0.88)
- Bayesian Adaptive Calibration and Optimal Design (NeurIPS 2024 · similarity 0.87)
- Nesting Particle Filters for Experimental Design in Dynamical Systems (ICML 2024 · similarity 0.87)
- PASOA- PArticle baSed Bayesian Optimal Adaptive design (ICML 2024 · similarity 0.86)
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.
