Research map / Probabilistic and scientific modeling / Bayesian optimization
“Why Not Looking backward?” A Robust Two-Step Method to Automatically Terminate Bayesian Optimization
Shuang Li, Ke Li, Wei Li
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 proposed backward-looking termination instead of Nguyen/Lorenz/Makarova thresholds | BO termination · Synthetic, RL and HPO benchmarks | performance degradation (Iperf): 0.0614 -> 0 | Improved |
| use proposed method instead of Naive no-change criterion | BO termination · Synthetic, RL and HPO benchmarks | relative computational cost (Icost): 1 -> 0.3012 | Improved |
| naive criterion versus proposed on performance degradation | BO termination · Synthetic, RL and HPO benchmarks | performance degradation (Iperf): 0 -> 0.0028 | Hurt |
| proposed method versus Nguyen and Lorenz on computational saving | BO termination · Synthetic, RL and HPO benchmarks | relative computational cost (Icost): 0.1048 -> 0.2209 | Hurt |
| proposed method versus Makarova at threshold eta1 | BO termination · Synthetic, RL and HPO benchmarks | relative computational cost (Icost): 0.7856 -> 0.6206 | Improved |
| relax the equality constraint of Condition 1 in the proposed method (ablation) | proposed termination · All benchmarks, eta_lb=2.05 | Icost and Iperf | Hurt |
| vary look-back window tau from 2 to 18 (ablation) | proposed termination · All benchmarks, eta_lb=2.05 | Icost and Iperf | No clear effect |
| use very small look-back window tau = 2 or 4 (ablation) | proposed termination · All benchmarks | Icost and Iperf | Hurt |
| use proposed criterion kappa_lb for threshold robustness across problems | proposed termination · Icdf across 9 benchmark problems | consistency of threshold | Improved |
Closest papers by meaning
- Stopping Bayesian Optimization with Probabilistic Regret Bounds (NeurIPS 2024 · similarity 0.89)
- Bayesian Optimisation with Unknown Hyperparameters: Regret Bounds Logarithmically Closer to Optimal (NeurIPS 2024 · similarity 0.89)
- This Too Shall Pass: Removing Stale Observations in Dynamic Bayesian Optimization (NeurIPS 2024 · similarity 0.89)
- BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability (NeurIPS 2026 · similarity 0.88)
- Optimizing the Unknown: Black Box Bayesian Optimization with Energy-Based Model and Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Principled Preferential Bayesian Optimization (ICML 2024 · 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.
