Research map / Probabilistic and scientific modeling / Bayesian optimization
Meta-Learning Acquisition Functions for Transfer Learning in Bayesian Optimization
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 |
|---|---|---|---|
| Meta-train a neural acquisition function on related objective functions | Bayesian optimization on translated and scaled Branin, Goldstein-Price, and Hartmann-3 · Global optimization benchmark functions | simple regret | Helped |
| Meta-train a neural acquisition function instead of using TAF | Bayesian optimization on translated and scaled Branin, Goldstein-Price, and Hartmann-3 · Global optimization benchmark functions | simple regret | Helped |
| Use the full available set of simulated source tasks rather than a selected subset (ablation) | MetaBO on Furuta pendulum stabilization · Simulation-to-real, simulation | simple regret | Helped |
| Transfer the simulation-trained neural acquisition function to the hardware without modification | MetaBO on Furuta pendulum stabilization · Simulation-to-real, hardware | optimization steps to find stabilizing controllers: less than ten | Helped |
| Use MetaBO instead of EI or TAF | Bayesian optimization of SVM and AdaBoost hyperparameters · Hyperparameter optimization | simple regret | Helped |
| Train on GP-sampled functions without coordinate inputs and apply the acquisition function across dimensions (ablation) | MetaBO on RBF- and Matern-5/2 GP function classes · General function classes, D=3, 4, and 5 | simple regret | No clear effect |
| Use a learned neural acquisition strategy instead of a source-prior combination with standard UCB or EI | MetaBO on Branin, Goldstein-Price, and Hartmann-3 · Global optimization benchmark functions | simple regret | Helped |
| Increase the number of TAF source tasks from 20 to 50 (ablation) | TAF on global optimization benchmark functions · Branin, Goldstein-Price, and Hartmann-3 | simple regret | Mixed |
| Use more MetaBO source tasks during training (ablation) | MetaBO on Branin · Branin | simple regret | No clear effect |
| Use more MetaBO source tasks during training (ablation) | MetaBO on Furuta pendulum stabilization · Simulation-to-real, simulation | simple regret | Helped |
Closest papers by meaning
- Learning search spaces for Bayesian optimization: Another view of hyperparameter transfer learning (NeurIPS 2019 · similarity 0.90)
- End-to-End Meta-Bayesian Optimisation with Transformer Neural Processes (NeurIPS 2023 · similarity 0.90)
- MALIBO: Meta-learning for Likelihood-free Bayesian Optimization (ICML 2024 · similarity 0.88)
- SMOG: Scalable Meta-Learning for Multi-Objective Bayesian Optimization (NeurIPS 2026 · similarity 0.87)
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.87)
- FunBO: Discovering Acquisition Functions for Bayesian Optimization with FunSearch (ICML 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.