Research map / Probabilistic and scientific modeling / Bayesian optimization
Batch Bayesian Optimization For Replicable Experimental Design
Zhongxiang Dai, Quoc Phong Nguyen, Sebastian Shenghong Tay, Daisuke Urano, Richalynn Leong, Bryan Kian Hsiang Low, Patrick Jaillet
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 |
|---|---|---|---|
| replicate each query a fixed number of times (n_t=1) instead of BTS-RED adaptive replication | Batch TS · Synthetic mean optimization (B=50) | simple regret | Hurt |
| use small fixed n_t (e.g. 5) versus large fixed n_t (e.g. 20) in Batch TS (ablation) | Batch TS · Synthetic mean optimization (B=50) | simple regret | Mixed |
| replace fixed replications with adaptive n_t chosen by BTS-RED-Known/Unknown | Batch TS · Synthetic mean optimization (B=50) | simple regret | Improved |
| set kappa=0.3 instead of kappa=0.2 in R^2 scaling (ablation) | BTS-RED · All experiments | simple regret / best value | Improved |
| vary kappa over a wide range (ablation) | BTS-RED · Synthetic mean optimization | simple regret | No clear effect |
| smaller kappa (0.2 vs 0.3) (ablation) | BTS-RED-Unknown · Synthetic mean optimization | average replications n_t | Improved |
| compare with sequential BO (GP-UCB, GP-TS, heteroscedastic GP-UCB) instead of batch methods | BTS-RED · Synthetic and precision agriculture mean optimization | simple regret | Hurt |
| use Mean-Var-BTS-RED instead of RAHBO (sequential, fixed replications) | Mean-variance optimization · Synthetic mean-variance (omega=0.3) | mean-variance simple regret | Improved |
| use BTS-RED-Unknown and Mean-Var-BTS-RED instead of Batch TS and RAHBO | Precision agriculture experiments · Leaf area and leaf area/tipburn combination | best observed value | Improved |
| use BTS-RED-Unknown and Mean-Var-BTS-RED with kappa 0.2 and 0.3 | SVM tuning on EMNIST · EMNIST SVM hyperparameter tuning (B=50) | best observed value | Improved |
| sequential RAHBO versus batch methods when noise variance is small | RAHBO · EMNIST SVM mean-variance | mean-variance objective | Hurt |
| vary budget B from 50 to larger or smaller (ablation) | BTS-RED · EMNIST SVM tuning | best observed value | No clear effect |
Closest papers by meaning
- Sequential Bayesian Experimental Design with Variable Cost Structure (NeurIPS 2020 · similarity 0.89)
- Bayesian Adaptive Calibration and Optimal Design (NeurIPS 2024 · similarity 0.87)
- Nesting Particle Filters for Experimental Design in Dynamical Systems (ICML 2024 · similarity 0.86)
- On Estimating the Gradient of the Expected Information Gain in Bayesian Experimental Design (AAAI 2024 · similarity 0.86)
- PASOA- PArticle baSed Bayesian Optimal Adaptive design (ICML 2024 · similarity 0.86)
- Bayesian Experimental Design Via Contrastive Diffusions (ICLR 2025 · 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.
