Gaussian Process Model for Estimating Piecewise Continuous Regression Functions

TL;DR

This paper introduces a novel Gaussian process model tailored for estimating piecewise continuous regression functions, effectively handling discontinuities by local data partitioning and boundary estimation, with demonstrated superior performance over traditional methods.

Contribution

The paper proposes a new GP modeling approach that estimates local regression functions with boundary-aware data partitioning, improving piecewise regression analysis.

Findings

Outperforms conventional GP methods in simulated tests

Efficient computation comparable to local GP approaches

Accurately estimates boundaries between data regions

Abstract

This paper presents a Gaussian process (GP) model for estimating piecewise continuous regression functions. In scientific and engineering applications of regression analysis, the underlying regression functions are piecewise continuous in that data follow different continuous regression models for different regions of the data with possible discontinuities between the regions. However, many conventional GP regression approaches are not designed for piecewise regression analysis. We propose a new GP modeling approach for estimating an unknown piecewise continuous regression function. The new GP model seeks for a local GP estimate of an unknown regression function at each test location, using local data neighboring to the test location. To accommodate the possibilities of the local data from different regions, the local data is partitioned into two sides by a local linear boundary, and only the local data belonging to the same side as the test location is used for the regression estimate. This local split works very well when the input regions are bounded by smooth boundaries, so the local linear approximation of the smooth boundaries works well. We estimate the local linear boundary jointly with the other hyperparameters of the GP model, using the maximum likelihood approach. Its computation time is as low as the local GP's time. The superior numerical performance of the proposed approach over the conventional GP modeling approaches is shown using various simulated piecewise regression functions.