Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Design Amortization for Bayesian Optimal Experimental Design
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Amortize a conditional variational model across designs rather than train a separate variational model for each design | Posterior estimator on the normal (linear) model · Normal (linear) model, 5 predictors, 5 experimental units | wall-clock time: 920s -> 293s | Helped |
| Use the design-amortized model instead of the non-amortized variational form | Posterior estimator on the normal (linear) model · 50 randomly generated designs | EIG approximation accuracy | Helped |
| Use the learned posterior in VNMC bounds rather than standard NMC | GLM model experiments · Normal, logistic, binomial, categorical, and multinomial GLMs | EIG bound tightness: 167× fewer samples | Helped |
| Train the variational posterior with the cheaper posterior estimator, then use it in VNMC | Design-amortized variational posterior · GLM model experiments | VNMC EIG estimation accuracy | Helped |
| Add attention layers to the set encoder instead of using residual layers (ablation) | Design-amortized posterior estimator · Linear unknown and binomial models, 5 predictors, prior covariance diagonal 25 | posterior-estimator loss | Helped |
| Add affine coupling transforms to the conditional base distribution (ablation) | Design-amortized posterior estimator · Linear unknown and binomial models, 5 predictors, prior covariance diagonal 25 | posterior-estimator loss | Helped |
| Increase affine coupling transforms from 4 to 8 (ablation) | Design-amortized posterior estimator · Linear unknown and binomial models, 5 predictors, prior covariance diagonal 25 | posterior-estimator loss | No clear effect |
| Replace affine coupling transforms with rational quadratic spline transforms (ablation) | Design-amortized posterior estimator · Linear unknown and binomial models, 5 predictors, prior covariance diagonal 25 | posterior-estimator loss | No clear effect |
Closest papers by meaning
- On Estimating the Gradient of the Expected Information Gain in Bayesian Experimental Design (AAAI 2024 · similarity 0.89)
- Bayesian Experimental Design Via Contrastive Diffusions (ICLR 2025 · similarity 0.88)
- Amortized Bayesian Experimental Design for Decision-Making (NeurIPS 2024 · similarity 0.88)
- Bayesian Adaptive Calibration and Optimal Design (NeurIPS 2024 · similarity 0.87)
- Maximin Robust Bayesian Experimental Design (NeurIPS 2026 · similarity 0.87)
- Nesting Particle Filters for Experimental Design in Dynamical Systems (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.