Research map / Probabilistic and scientific modeling / Bayesian optimization
Few-Shot Bayesian Optimization with Deep Kernel Surrogates
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 |
|---|---|---|---|
| Use few-shot meta-learned deep-kernel GP optimization instead of the strongest reported baseline | Bayesian optimization · AdaBoost, 50 trials | normalized regret: 1.02 -> 0.80 | Helped |
| Use few-shot meta-learned deep-kernel GP optimization instead of the strongest reported baseline | Bayesian optimization · GLMNet, 100 trials | normalized regret: 0.24 -> 0.16 | Helped |
| Use few-shot meta-learned deep-kernel GP optimization instead of the strongest reported baseline | Bayesian optimization · SVM, 100 trials | normalized regret: 0.65 -> 0.36 | Helped |
| Add target-task fine-tuning of the deep-kernel parameters (ablation) | Multi-Head GP · AdaBoost, GLMNet, and SVM | normalized regret | Helped |
| Replace separate task-specific Gaussian processes with one shared Gaussian process (ablation) | Multi-Head GP · AdaBoost, GLMNet, and SVM | normalized regret | Helped |
| Add data-driven warm-start initialization (ablation) | Bayesian optimization · GLMNet and SVM | normalized regret | Helped |
| Add data-driven warm-start initialization (ablation) | Bayesian optimization · AdaBoost | normalized regret | No clear effect |
| Use Reptile instead of the paper's meta-learning approach (ablation) | FSBO · AdaBoost, GLMNet, and SVM | normalized regret: AdaBoost: 0.80 -> 1.21; GLMNet: 0.16 -> 0.14; SVM: 0.36 -> 0.72 | Mixed |
| Train MetaBO for longer trial episodes rather than the shorter setting (ablation) | MetaBO · AdaBoost, GLMNet, and SVM | normalized regret | Hurt |
Closest papers by meaning
- Meta-Surrogate Benchmarking for Hyperparameter Optimization (NeurIPS 2019 · similarity 0.91)
- Supervising the Multi-Fidelity Race of Hyperparameter Configurations (NeurIPS 2022 · similarity 0.89)
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.86)
- Cost-Sensitive Freeze-thaw Bayesian Optimization for Efficient Hyperparameter Tuning (NeurIPS 2025 · similarity 0.85)
- Informed Initialization for Bayesian Optimization and Active Learning (NeurIPS 2025 · similarity 0.85)
- Efficient Hyperparameter Optimization with Adaptive Fidelity Identification (CVPR 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.