Research map / Probabilistic and scientific modeling / Bayesian optimization
Multi-Step Budgeted Bayesian Optimization with Unknown Evaluation Costs
Raul Astudillo, Daniel R. Jiang, Maximilian Balandat, Eytan Bakshy, Peter I. Frazier
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 4 look-ahead steps instead of 2 in B-MS-EI (ablation) | B-MS-EI · all benchmark problems | log-regret | Improved |
| replace cost-agnostic EI with B-MS-EI (4 steps) | BO acquisition · Dropwave, CNN | log-regret | Improved |
| replace EI-PUC / EI-PUC-CC with B-MS-EI (4 steps) | BO acquisition · Alpine, Shekel, LDA | log-regret | Improved |
| use multiple fantasy samples per step (4-B-MS-EI) instead of a single path (4-B-MS-EI_p) (ablation) | B-MS-EI · runtime per acquisition, Dropwave | seconds: 42.6 -> 87.2 | Hurt |
| full tree vs path variant, Alpine1 runtime (ablation) | B-MS-EI · Alpine1 | seconds: 106.1 -> 236.7 | Hurt |
| full tree vs path variant, CNN runtime (ablation) | B-MS-EI · CNN | seconds: 286.1 -> 443.2 | Hurt |
| full tree vs path variant, Ackley runtime (ablation) | B-MS-EI · Ackley | seconds: 95.8 -> 103.5 | Hurt |
| increase look-ahead from 2 to 4 steps (runtime), Dropwave (ablation) | B-MS-EI · Dropwave | seconds: 26.7 -> 87.2 | Hurt |
| increase look-ahead from 2 to 4 steps (runtime), CNN (ablation) | B-MS-EI · CNN | seconds: 95.0 -> 443.2 | Hurt |
| use B-MS-EI instead of EI (runtime), CNN | BO acquisition · CNN | seconds: 11.7 -> 443.2 | Hurt |
| use B-MS-EI instead of EI (runtime), Boston | BO acquisition · Boston | seconds: 1.8 -> 90.5 | Hurt |
| halve the budget for B-MS-EI (ablation) | B-MS-EI · Dropwave, LDA, Robot Pushing | log-regret | Improved |
Closest papers by meaning
- Bayesian Optimization for Unknown Cost-Varying Variable Subsets with No-Regret Costs (AAAI 2025 · similarity 0.90)
- Cost-aware Bayesian Optimization via the Pandora's Box Gittins Index (NeurIPS 2024 · similarity 0.89)
- Informed Initialization for Bayesian Optimization and Active Learning (NeurIPS 2025 · similarity 0.89)
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.89)
- Optimizing the Unknown: Black Box Bayesian Optimization with Energy-Based Model and Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.88)
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.
