Efficiently Computable Safety Bounds for Gaussian Processes in Active Learning

被引:0
|
作者
Tebbe, Joern [1 ]
Zimmer, Christoph [2 ]
Steland, Ansgar [3 ]
Lange-Hegermann, Markus [1 ]
机构
[1] OWL Univ Appl Sci & Arts, Lemgo, Germany
[2] Bosch Ctr Artificial Intelligence, Renningen, Germany
[3] Rhein Westfal TH Aachen, Aachen, Germany
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Active learning of physical systems must commonly respect practical safety constraints, which restricts the exploration of the design space. Gaussian Processes (GPs) and their calibrated uncertainty estimations are widely used for this purpose. In many technical applications the design space is explored via continuous trajectories, along which the safety needs to be assessed. This is particularly challenging for strict safety requirements in GP methods, as it employs computationally expensive Monte-Carlo sampling of high quantiles. We address these challenges by providing provable safety bounds based on the adaptively sampled median of the supremum of the posterior GP. Our method significantly reduces the number of samples required for estimating high safety probabilities, resulting in faster evaluation without sacrificing accuracy and exploration speed. The effectiveness of our safe active learning approach is demonstrated through extensive simulations and validated using a real-world engine example.
引用
收藏
页数:21
相关论文
共 50 条
  • [41] Labeling Examples That Matter: Relevance-Based Active Learning with Gaussian Processes
    Freytag, Alexander
    Rodner, Erik
    Bodesheim, Paul
    Denzler, Joachim
    [J]. PATTERN RECOGNITION, GCPR 2013, 2013, 8142 : 282 - 291
  • [42] UPPER BOUNDS FOR ASYMPTOTIC MAXIMA OF CONTINUOUS GAUSSIAN PROCESSES
    MARCUS, MB
    [J]. ANNALS OF MATHEMATICAL STATISTICS, 1972, 43 (02): : 522 - &
  • [43] A Receding Horizon Approach for Simultaneous Active Learning and Control using Gaussian Processes
    Le, Viet-Anh
    Nghiem, Truong X.
    [J]. 5TH IEEE CONFERENCE ON CONTROL TECHNOLOGY AND APPLICATIONS (IEEE CCTA 2021), 2021, : 453 - 458
  • [44] An Active Learning Based Robot Kinematic Calibration Framework Using Gaussian Processes
    Das, Ersin
    Burdick, Joel W.
    [J]. 2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023), 2023, : 11495 - 11501
  • [45] Assessment and treatment of visuospatial neglect using active learning with Gaussian processes regression
    De Boi, Ivan
    Embrechts, Elissa
    Schatteman, Quirine
    Penne, Rudi
    Truijen, Steven
    Saeys, Wim
    [J]. ARTIFICIAL INTELLIGENCE IN MEDICINE, 2024, 149
  • [46] Active learning-assisted neutron spectroscopy with log-Gaussian processes
    Mario Teixeira Parente
    Georg Brandl
    Christian Franz
    Uwe Stuhr
    Marina Ganeva
    Astrid Schneidewind
    [J]. Nature Communications, 14
  • [47] Learning curves for Gaussian processes
    Sollich, P
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 11, 1999, 11 : 344 - 350
  • [48] Gaussian processes in machine learning
    Rasmussen, CE
    [J]. ADVANCED LECTURES ON MACHINE LEARNING, 2004, 3176 : 63 - 71
  • [49] Lifelong Learning with Gaussian Processes
    Clingerman, Christopher
    Eaton, Eric
    [J]. MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2017, PT II, 2017, 10535 : 690 - 704
  • [50] Learning Gradients with Gaussian Processes
    Jiang, Xinwei
    Gao, Junbin
    Wang, Tianjiang
    Kwan, Paul W.
    [J]. ADVANCES IN KNOWLEDGE DISCOVERY AND DATA MINING, PT II, PROCEEDINGS, 2010, 6119 : 113 - +