Research map / Optimization / Optimizers
Frugal Optimization for Cost-related Hyperparameters
Qingyun Wu, Chi Wang, Silu Huang
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 cost-frugal randomized direct search instead of the compared HPO methods | XGBoost hyperparameter tuning · Airlines | scaled score: 1.3996; existing RS and BO methods report more than 16% lower score | Improved |
| Use cost-frugal randomized direct search instead of other methods | XGBoost hyperparameter tuning · shuttle | time to reach scaled score 0.9999: 72 seconds; other methods show 6× to 26× slowdowns | Improved |
| Replace OPT with ZOGD in the HPO framework (ablation) | HPO · XGBoost benchmarks | anytime performance | Hurt |
| Penalize expensive configurations using expected improvement per second | GPEI · adult and car | validation loss over time | Mixed |
| Use HPO rather than the compared baselines | XGBoost hyperparameter tuning · XGBoost benchmarks | fraction reaching best loss: 96%; baselines less than 19% | Improved |
| Use HPO rather than the compared baselines | XGBoost hyperparameter tuning · XGBoost benchmarks | normalized score: There are only two cases where HPO's score is lower than a baseline, i.e., BOHB, by 0.4% | Mixed |
| Use randomized restarts when progress stalls (ablation) | HPO · XGBoost and DNN benchmarks | validation loss | No clear effect |
| Adapt the step size based on search progress (ablation) | HPO · XGBoost and DNN benchmarks | validation loss | No clear effect |
| Project proposed configurations into the feasible hyperparameter space (ablation) | HPO · XGBoost and DNN benchmarks | validation loss | No clear effect |
Closest papers by meaning
- HPO-B: A Large-Scale Reproducible Benchmark for Black-Box HPO based on OpenML (NeurIPS 2021 · similarity 0.91)
- Interactive Hyperparameter Optimization in Multi-Objective Problems via Preference Learning (AAAI 2024 · similarity 0.88)
- Hyperparameter Optimization via Sequential Uniform Designs (JMLR 2021 · similarity 0.88)
- Meta-Surrogate Benchmarking for Hyperparameter Optimization (NeurIPS 2019 · similarity 0.87)
- UQ-Guided Hyperparameter Optimization for Iterative Learners (NeurIPS 2024 · similarity 0.86)
- Cost-Sensitive Freeze-thaw Bayesian Optimization for Efficient Hyperparameter Tuning (NeurIPS 2025 · 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.
