Research map / Probabilistic and scientific modeling / Bayesian optimization
An Empirical Study of Bayesian Optimization: Acquisition Versus Partition
Erich Merrill, Alan Fern, Xiaoli Fern, Nima Dolatnia
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 EI instead of GP-UCB acquisition | ABO · 23 benchmark functions, t=500 | pairwise wins-losses-ties: 12 - 3 - 8 | Improved |
| use ABO-EI instead of PGO-SOO | ABO-EI · 23 benchmark functions, t=500 | pairwise wins-losses-ties: 18 - 1 - 4 | Improved |
| add GP posterior to SOO (BaMSOO) | PGO-SOO · 23 benchmark functions, t=500 | pairwise wins-losses-ties: PBO-BaMSOO vs PGO-SOO 10 - 1 - 12 reported as PBO losing once | Improved |
| use DIRECT instead of SOO | PGO-SOO · 23 benchmark functions, t=500 | pairwise wins-losses-ties: 13 - 0 - 10 | Improved |
| give PGO 10,000 samples instead of 500 | SOO and LOGO · 23 benchmark functions | pairwise wins vs ABO-EI: 18 - 1 - 4 to 13 - 4 - 6 | Mixed |
| give DIRECT 10,000 samples instead of 500 | DIRECT · 23 benchmark functions | pairwise vs ABO-EI: 3 - 1 - 19 to 0 - 9 - 14 | Improved |
| give PBO the same wall-clock time as ABO (thousands of extra samples) | PBO-BaMSOO and PBO-IMGPO · Branin, Rosenbrock 6, Schwefel 10 | regret | No clear effect |
| update GP hyperparameters every sixteen observations instead of every second (sparse ABO) (ablation) | ABO-EI · Branin, Rosenbrock 6, Schwefel 10 | regret | No clear effect |
| tune UCB beta (constants 0.1, 1.0, 5.0, 10.0, schedules) (ablation) | ABO-UCB · Ackley 4, Rastrigin 6, Rosenbrock 10 | regret | Mixed |
| use Kandasamy schedule UCB-K instead of Krause beta schedule (ablation) | ABO-UCB · Ackley 4, Rastrigin 6, Rosenbrock 10 | regret | Improved |
| restrict to 2-D functions rather than D>2 | ABO-EI vs PGO · D=2 functions | pairwise wins-losses-ties: D=2: 3 - 1 - 2 vs LOGO; D>2: 15 - 0 - 2 | Mixed |
Closest papers by meaning
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.91)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.90)
- BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability (NeurIPS 2026 · similarity 0.90)
- A General Framework for User-Guided Bayesian Optimization (ICLR 2024 · similarity 0.89)
- Principled Preferential Bayesian Optimization (ICML 2024 · similarity 0.89)
- Simple Extensions of Single-Objective Acquisition Functions and Hedge Strategies for Multi-Objective Bayesian Optimization (NeurIPS 2026 · similarity 0.89)
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.
