Research map / Probabilistic and scientific modeling / Bayesian optimization
BOHB: Robust and Efficient Hyperparameter Optimization at Scale
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Combine Hyperband's multi-fidelity successive halving with model-guided TPE sampling | Hyperparameter optimization · OpenML feed-forward neural network surrogates | incumbent performance: 100 times faster | Helped |
| Use model-guided configuration sampling instead of random sampling in Hyperband | Hyperband · Counting Ones, 16 dimensions | immediate regret | Helped |
| Use multi-fidelity evaluation with Hyperband | Bayesian optimization · Feed-forward neural network surrogates | incumbent performance: roughly three-fold speedup | Helped |
| Use BOHB rather than TPE | Hyperparameter optimization · SVM on MNIST surrogate | immediate regret | No clear effect |
| Use BOHB rather than Hyperband | Hyperparameter optimization · SVM on MNIST surrogate | immediate regret | Helped |
| Use BOHB rather than Hyperband or TPE | Bayesian neural network hyperparameter optimization · Boston housing | negative log-likelihood | Helped |
| Use BOHB rather than Hyperband | PPO hyperparameter optimization · Cartpole swing-up | episodes until convergence | Helped |
| Use a 16-dimensional search space rather than a smaller-dimensional space | BOHB · Counting Ones, 64 dimensions | immediate regret | Hurt |
| Use BOHB's shared worker pool and prioritize smaller-budget runs | BOHB parallelization · Letter surrogate benchmark | speedup to achieve regret of 10^-2: 2000s / 130s ≈ 15 | Helped |
| Increase workers from one to 32 | BOHB parallelization · Letter surrogate benchmark | speedup to achieve regret of 10^-2: 2000s / 130s ≈ 15 | Helped |
Closest papers by meaning
- Hyperband: Bandit-Based Configuration Evaluation for Hyperparameter Optimization (ICLR 2017 · similarity 0.91)
- Efficient Hyperparameter Optimization with Adaptive Fidelity Identification (CVPR 2024 · similarity 0.89)
- Bayesian Optimization Meets Bayesian Optimal Stopping (ICML 2019 · similarity 0.89)
- Learning search spaces for Bayesian optimization: Another view of hyperparameter transfer learning (NeurIPS 2019 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.88)
- Cost-Sensitive Freeze-thaw Bayesian Optimization for Efficient Hyperparameter Tuning (NeurIPS 2025 · similarity 0.87)
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.