Research map / Probabilistic and scientific modeling / Bayesian optimization
A General Recipe for Likelihood-free Bayesian Optimization
Jiaming Song, Lantao Yu, Willie Neiswanger, Stefano Ermon
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 LFBO (EI) instead of the compared black-box optimization methods | HPOBench neural-network tuning · Protein, Slice, Naval, Parkinsons | immediate regret | Improved |
| Use LFBO (EI) instead of the compared black-box optimization methods | NAS-Bench-201 · CIFAR-100 and ImageNet-16 | immediate regret | Improved |
| Use SMAC instead of LFBO (EI) | NAS-Bench-201 · CIFAR-10 | immediate regret | Improved |
| Use EI utility rather than PI utility | LFBO with MLP classifier · HPOBench: Naval | immediate regret | No clear effect |
| Use EI utility rather than PI utility | LFBO with MLP, RF, or XGBoost classifier · HPOBench and NAS-Bench-201, excluding Naval with MLP | immediate regret | Improved |
| Use GP-based EI instead of LFBO (EI) | HPOBench and NAS-Bench-201 · real-world benchmark tasks | immediate regret | Hurt |
| Use utility exponent λ = 1.5 instead of EI's λ = 1 (ablation) | LFBO with XGBoost · HPOBench: two problems | optimization performance | Improved |
| Increase the utility exponent from λ = 0 (PI) toward λ = 1 (EI) (ablation) | LFBO with XGBoost · HPOBench: two problems | optimization performance | Improved |
| Increase γ (ablation) | LFBO (EI) with XGBoost · HPOBench: two problems | optimization performance | Hurt |
| Increase γ rather than use the smaller threshold setting (ablation) | LFBO (EI) with XGBoost · HPOBench: two problems | optimization performance | Mixed |
| Incorporate the known composite objective structure into LFBO | chemical-accident environmental model · composite objective optimization | immediate regret | Improved |
| Replace the structure-aware final layer with another neural network (ablation) | Composite LFBO (EI) · composite objective optimization | optimization performance | Hurt |
Closest papers by meaning
- Simple Extensions of Single-Objective Acquisition Functions and Hedge Strategies for Multi-Objective Bayesian Optimization (NeurIPS 2026 · similarity 0.89)
- BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability (NeurIPS 2026 · similarity 0.89)
- Principled Preferential Bayesian Optimization (ICML 2024 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.89)
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.89)
- Stochastic Bayesian Optimization with Unknown Continuous Context Distribution via Kernel Density Estimation (AAAI 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.
