Robust Gaussian process regression with a bias model
作者：
Highlights：
• This paper presents a Gaussian process regression approach that provides the regression outcomes robust to outliers.
• The proposed approach models an outlier as a noisy and biased observation of an unknown regression function.
• Two bias models are presented to model outliers.
• The ML estimation of the proposed models is much more computationally efficient and accurate than the existing MCMC-based approaches.
• The approach was validated using a comprehensive simulation study and the application to environmental data analysis.
摘要
•This paper presents a Gaussian process regression approach that provides the regression outcomes robust to outliers.•The proposed approach models an outlier as a noisy and biased observation of an unknown regression function.•Two bias models are presented to model outliers.•The ML estimation of the proposed models is much more computationally efficient and accurate than the existing MCMC-based approaches.•The approach was validated using a comprehensive simulation study and the application to environmental data analysis.
论文关键词：Robust regression,Gaussian process,Random bias estimation,Regularized likelihood maximization,Sensor data
论文评审过程：Received 13 August 2020, Revised 9 August 2021, Accepted 18 November 2021, Available online 21 November 2021, Version of Record 28 February 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108444
