Robust Gaussian Process Regression Based on Iterative Trimming

TL;DR

This paper introduces a robust Gaussian process regression method that iteratively trims outliers, significantly improving accuracy on contaminated data and demonstrating practical effectiveness in astrophysical applications.

Contribution

The paper proposes a new robust GP regression algorithm based on iterative trimming, offering improved accuracy and simplicity over existing robust GP methods.

Findings

Outperforms standard GP and Student-t likelihood-based methods in contaminated data scenarios.

Effectively determines the main-sequence ridge line in star cluster diagrams.

Easier to implement than previous robust GP variants.

Abstract

The Gaussian process (GP) regression can be severely biased when the data are contaminated by outliers. This paper presents a new robust GP regression algorithm that iteratively trims the most extreme data points. While the new algorithm retains the attractive properties of the standard GP as a nonparametric and flexible regression method, it can greatly improve the model accuracy for contaminated data even in the presence of extreme or abundant outliers. It is also easier to implement compared with previous robust GP variants that rely on approximate inference. Applied to a wide range of experiments with different contamination levels, the proposed method significantly outperforms the standard GP and the popular robust GP variant with the Student-t likelihood in most test cases. In addition, as a practical example in the astrophysical study, we show that this method can precisely determine the main-sequence ridge line in the color-magnitude diagram of star clusters.