Research map / Probabilistic and scientific modeling / Bayesian optimization
Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization
Lisha Li, Kevin Jamieson, Giulia DeSalvo, Afshin Rostamizadeh, Ameet Talwalkar
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 Hyperband instead of random search/Bayesian optimization | deep-network tuning · CIFAR-10 | speedup: over 20x faster than random search | Improved |
| use Hyperband instead of SMAC (early) | deep-network tuning · MRBI | speedup: 5x faster than SMAC (early) | Improved |
| use Hyperband instead of SMAC (early) | deep-network tuning · SVHN | test error | Hurt |
| repeat only the most aggressive bracket s=4 instead of full Hyperband (ablation) | Hyperband · CIFAR-10 and MRBI | test error | Improved |
| use infinite horizon instead of finite horizon Hyperband (ablation) | Hyperband · CIFAR-10 | test error within 50R | Hurt |
| use Hyperband with data subsampling instead of random search | Hyperband · 117 datasets | test error rank | Mixed |
| use random 2x instead of Hyperband | random search · 117 datasets | test error rank | Improved |
| use Hyperband instead of other searchers on selected datasets | Hyperband · 21 of 117 datasets | test error rank | Improved |
| use Hyperband instead of Bayesian optimization | kernel classifier · Kernel least squares CIFAR-10 | speedup: over 30x faster than Bayesian optimization, 70x faster than random search | Improved |
| use Hyperband instead of random search | random feature kernel · Random feature kernel approximation | speedup: around 6x faster | Improved |
| use Hyperband instead of random search with brackets s=0 | Hyperband · Appendix LeNet (R=81, eta=3) | test error | Improved |
Later papers that compared with this method
3 later accepted papers report Hyperband as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Hyperparameter Optimization via Sequential Uniform Designs (JMLR 2021 · better on 8 of 8 shared results)
- AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly (ICLR 2021 · better on 2 of 2 shared results)
- Tuning Hyperparameters without Grad Students: Scalable and Robust Bayesian Optimisation with Dragonfly (JMLR 2020 · better on 0 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Hyperband: Bandit-Based Configuration Evaluation for Hyperparameter Optimization (ICLR 2017 · similarity 0.97)
- BOHB: Robust and Efficient Hyperparameter Optimization at Scale (ICML 2018 · similarity 0.91)
- Put CASH on Bandits: A Max K-Armed Problem for Automated Machine Learning (NeurIPS 2025 · similarity 0.87)
- Offline-to-Online Hyperparameter Transfer for Stochastic Bandits (AAAI 2025 · similarity 0.86)
- Random Exploration in Bayesian Optimization: Order-Optimal Regret and Computational Efficiency (ICML 2024 · similarity 0.85)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · 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.
