Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Fast Bayesian Inference with Batch Bayesian Quadrature via Kernel Recombination
Masaki Adachi, Satoshi Hayakawa, Martin Jørgensen, Harald Oberhauser, Michael A Osborne
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 |
|---|---|---|---|
| add alternate kernel update and uncertainty sampling (vs. removing kernel update) (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -1.5936 -> -4.0138 | Improved |
| add factorisation trick (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | wall time (s): 367.63 -> 48.75 | Improved |
| use linear IVR instead of optimal (square-root) IVR (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | wall time (s): 810.45 -> 48.75 | Improved |
| use linear IVR instead of optimal IVR (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -3.4379 -> -4.0138 | Improved |
| add uncertainty sampling and IVR proposal on top of factorisation trick (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -3.3310 -> -4.0138 | Improved |
| add RCHQ with all other components versus kernel-update-only baseline (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | wall time (s): 407.42 -> 48.75 | Improved |
| replace WSABI-L with WSABI-M (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -4.0138 -> -4.0418 | No clear effect |
| replace WSABI-L with WSABI-M (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | wall time (s): 48.75 -> 814.13 | Hurt |
| replace WSABI-L with vanilla BQ (no warp) (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logKL: -9.6222 -> 12.307 | Hurt |
| replace WSABI-L with BBQ (log warp) (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -4.0138 -> -3.1278 | Hurt |
| use Exponential kernel instead of RBF (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logMAE: -2.9598 -> -4.6094 | Improved |
| use Exponentiated quadratic kernel instead of RBF (ablation) | BASQ · 10-D Gaussian mixture, 1000 observations, batch 100 | logKL: 12.321 -> 11.046 | Improved |
Closest papers by meaning
- Automated Model Selection with Bayesian Quadrature (ICML 2019 · similarity 0.90)
- Bayesian Probabilistic Numerical Integration with Tree-Based Models (NeurIPS 2020 · similarity 0.87)
- Convergence Guarantees for Adaptive Bayesian Quadrature Methods (NeurIPS 2019 · similarity 0.87)
- Consistency Models for Scalable and Fast Simulation-Based Inference (NeurIPS 2024 · similarity 0.85)
- Effortless, Simulation-Efficient Bayesian Inference using Tabular Foundation Models (NeurIPS 2025 · similarity 0.85)
- Provable Posterior Sampling with Denoising Oracles via Tilted Transport (NeurIPS 2024 · similarity 0.85)
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.
