mlsanityData to 14 Oct 2026

Meta-Learning Priors for Safe Bayesian Optimization

arXiv 2210.00762 ↗First posted 3 Oct 2022cs.LG cs.AI cs.ROcited by 14

Not on the map. Why?

Abstract (arXiv)

In robotics, optimizing controller parameters under safety constraints is an important challenge. Safe Bayesian optimization (BO) quantifies uncertainty in the objective and constraints to safely guide exploration in such settings. Hand-designing a suitable probabilistic model can be challenging, however. In the presence of unknown safety constraints, it is crucial to choose reliable model hyper-parameters to avoid safety violations. Here, we propose a data-driven approach to this problem by meta-learning priors for safe BO from offline data. We build on a meta-learning algorithm, F-PACOH, capable of providing reliable uncertainty quantification in settings of data scarcity. As core contribution, we develop a novel framework for choosing safety-compliant priors in a data-riven manner via empirical uncertainty metrics and a frontier search algorithm. On benchmark functions and a high-precision motion system, we demonstrate that our meta-learned priors accelerate the convergence of safe BO approaches while maintaining safety.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Builds on 36 of its 59 references that are in the map. First posted on arXiv 3 Oct 2022.

Builds on

36 of 59 references.

References · 36

  1. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,897
  2. 2017Prototypical Networks for Few-shot Learningcited by 3,015
  3. 2016Matching Networks for One Shot Learningcited by 2,255
  4. 2018On First-Order Meta-Learning Algorithmscited by 1,018
  5. 2018A Tutorial on Bayesian Optimizationcited by 744
  6. 2016Learning to learn by gradient descent by gradient descentcited by 620
  7. 2017Ray: A Distributed Framework for Emerging AI Applicationscited by 527
  8. 2015Deep Kernel Learningcited by 445
  9. 2018Accurate Uncertainties for Deep Learning Using Calibrated Regressioncited by 422
  10. 2017Max-value Entropy Search for Efficient Bayesian Optimizationcited by 206
  11. 2018Bayesian Model-Agnostic Meta-Learningcited by 194
  12. 2016Bayesian Optimization with Safety Constraints: Safe and Automatic Parameter Tuning in Roboticscited by 134
24 more
  1. 2019Meta-Learning without Memorizationcited by 92
  2. 2018ProMP: Proximal Meta-Policy Searchcited by 91
  3. 2020PACOH: Bayes-Optimal Meta-Learning with PAC-Guaranteescited by 85
  4. 2018Stagewise Safe Bayesian Optimization with Gaussian Processescited by 79
  5. 2015Safe Controller Optimization for Quadrotors with Gaussian Processescited by 78
  6. 2013A Literature Survey of Benchmark Functions For Global Optimization Problemscited by 77
  7. 2017Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theorycited by 76
  8. 2013A PAC-Bayesian bound for Lifelong Learningcited by 69
  9. 2015Predictive Entropy Search for Bayesian Optimization with Unknown Constraintscited by 51
  10. 2019Safe Exploration for Interactive Machine Learningcited by 51
  11. 2018Meta-Learning Priors for Efficient Online Bayesian Regressioncited by 48
  12. 2016Automatic LQR Tuning Based on Gaussian Process Global Optimizationcited by 30
  13. 2021Generalization Bounds for Meta-Learning via PAC-Bayes and Uniform Stabilitycited by 30
  14. 2019Safe Contextual Bayesian Optimization for Sustainable Room Temperature PID Control Tuningcited by 24
  15. 2021Meta-Learning Reliable Priors in the Function Spacecited by 23
  16. 2018Data-efficient Auto-tuning with Bayesian Optimization: An Industrial Control Studycited by 15
  17. 2019AI-optimized detector design for the future Electron-Ion Collider: the dual-radiator RICH casecited by 15
  18. 2020Safety-Aware Cascade Controller Tuning Using Constrained Bayesian Optimizationcited by 15
  19. 2021Safe and Efficient Model-free Adaptive Control via Bayesian Optimizationcited by 10
  20. 2020The Murphy Decomposition and the Calibration-Resolution Principle: A New Perspective on Forecast Evaluationcited by 8
  21. 2021Performance-based Trajectory Optimization for Path Following Control Using Bayesian Optimizationcited by 8
  22. 2022Meta Representation Learning with Contextual Linear Banditscited by 6
  23. 2022Meta-Learning Hypothesis Spaces for Sequential Decision-makingcited by 5
  24. 2020Cascade Control: Data-Driven Tuning Approach Based on Bayesian Optimizationcited by 4

Built on by

Cited by year

Citing papers per year: 2022: 3; 2023: 5; 2024: 6.

14 papers.

Sort:

  1. 2022Scalable PAC-Bayesian Meta-Learning via the PAC-Optimal Hyper-Posterior: From Theory to Practicecited by 11
  2. 2023Bayesian Optimization with Formal Safety Guarantees via Online Conformal Predictioncited by 10
  3. 2023Tuning Legged Locomotion Controllers via Safe Bayesian Optimizationcited by 9
  4. 2023Data-Efficient Task Generalization via Probabilistic Model-based Meta Reinforcement Learningcited by 6
  5. 2022MARS: Meta-Learning as Score Matching in the Function Spacecited by 5
  6. 2022Lifelong Bandit Optimization: No Prior and No Regretcited by 4
  7. 2023Scalable Meta-Learning with Gaussian Processescited by 3
  8. 2024Multi-Fidelity Bayesian Optimization With Across-Task Transferable Max-Value Entropy Searchcited by 3
  9. 2024Active Few-Shot Fine-Tuningcited by 2
  10. 2024Global Safe Sequential Learning via Efficient Knowledge Transfercited by 1
  11. 2024Robust Bayesian Optimization via Localized Online Conformal Predictioncited by 1
  12. 2023META-SMGO-Delta: similarity as a prior in black-box optimizationcited by 0
2 more
  1. 2024Streaming Gaussian Dirichlet Random Fields for Spatial Predictions of High Dimensional Categorical Observationscited by 0
  2. 2024Safe Bayesian Optimization for Complex Control Systems via Additive Gaussian Processescited by 0