Bayesian Optimization for Performance Auto-Tuning

    In many engineering systems, the performance highly depends on some hand-tuned parameters. The mapping from these parameters to the performance is typically a black-box/grey-box function. Meanwhile, there are also usually black-box closed-loop constraints that need to be taken care of. In this project, we aim to develop sample-efficient black-box optimization method to tune these parameters, with formal guarantees on running performance and violations. We also release solver to facilitate the real-world applications of our methods.

Publications: Theory

  1. “Lower Bounds on the Noiseless Worst-Case Complexity of Efficient Global Optimization” [PDF]
    Wenjie Xu, Yuning Jiang, Emilio T. Maddalena, Colin N. Jones, accepted to Journal of Optimization Theory and Applications (JOTA), 2024.

Publications: Algorithms

  1. “Principled Preferential Bayesian Optimization” [arXiv]
    Wenjie Xu, Wenbin Wang, Yuning Jiang, Bratislav Svetozarevic, and Colin N. Jones, accepted to the Forty-first International Conference on Machine Learning (ICML 2024, Oral, top 1.5%).

  2. “Primal-Dual Contextual Bayesian Optimization for Control System Online Optimization with Time-Average Constraints” [PDF] [Technical Report]
    Wenjie Xu, Yuning Jiang, Bratislav Svetozarevic, Colin N. Jones, accepted to the 62nd IEEE Conference on Decision and Control (CDC 2023).

  3. “Constrained Efficient Global Optimization of Expensive Black-box Functions” [PDF] [Video] [Poster]
    Wenjie Xu, Yuning Jiang, Bratislav Svetozarevic, and Colin N. Jones, accepted to the Fortieth International Conference on Machine Learning (ICML 2023).

Publications: Applications

  1. “Violation-Aware Contextual Bayesian Optimization for Controller Performance Optimization with Unmodeled Constraints” [arXiv]
    Wenjie Xu, Colin N Jones, Bratislav Svetozarevic, Christopher R. Laughman, and Ankush Chakrabarty, accepted to Journal of Process Control (JPC), 2024.

  2. “ Data-driven adaptive building thermal controller tuning with constraints: A primal-dual contextual Bayesian optimization approach” [arXiv]
    Wenjie Xu, Bratislav Svetozarevic, Loris Di Natale, Philipp Heer, Colin N Jones, accepted to Applied Energy (APEN), 2023.

  3. CONFIG: Constrained Efficient Global Optimization for Closed-Loop Control System Optimization with Unmodeled Constraints” [PDF] [Code]
    Wenjie Xu, Yuning Jiang, Bratislav Svetozarevic, Philipp Heer, and Colin N. Jones, accepted to IFAC World Congress 2023. (IFAC 2023)

  4. “VABO: Violation-Aware Bayesian Optimization for Closed-Loop Control Performance Optimization with Unmodeled Constraints” [PDF] [Video] [Code]
    Wenjie Xu, Colin N Jones, Bratislav Svetozarevic, Christopher R. Laughman, and Ankush Chakrabarty, accepted to the 2022 American Control Conference (ACC 2022).

    (ASME Energy Systems Technical Committee Best Paper Award)

Preprints

  1. “ Multi-Agent Bayesian Optimization with Coupled Black-Box and Affine Constraints” [arXiv]
    Wenjie Xu, Yuning Jiang, Bratislav Svetozarevic, Colin N. Jones.

  2. “ Bayesian Optimization of Expensive Nested Grey-Box Functions ” [arXiv]
    Wenjie Xu, Yuning Jiang, Bratislav Svetozarevic, Colin N. Jones.

Software

  1. [POP-BO: Principled Optimistic Preferential Bayesian Optimization]
  2. [VABO: Violation-Aware Bayesian Optimization]
  3. [CONFIG: CONstrained efFIcient Global Optimization toolbox]