Reduced Rank Gaussian Process Regression (GPR)

This repository provides example runs of reduced-rank Gaussian process regression (GPR) and kernel function approximation.

Configuration

You can set hyperparameters for the kernel function (currently, only the Matern kernel is available) and parameters for the Hilbert space approximation method in 'config.yaml'.

Example Scripts

• 'Kern_apx_example.py': Provides comparison results between the original kernel function and the approximated kernel function.

• 'Estimation_example.py': Provides comparison results between standard kernel-based GPR and reduced-rank GPR using a 2-D sample scalar function.

Library

• 'ReducedRankGPLib.py': A library for sparse GPR, including methods for model updates, query test inputs, and more.

requirements for kernel-based GPR

pip3 install GPy