GMR-based Gaussian Process

This folder contains the different toy examples presented in the paper "Learning from Demonstration with model-based Gaussian Processes" (N. Jaquier, D. Ginsbourger and S. Calinon), CoRL 2019.

Installation

These examples work with Python 2 and 3. First install the following packages:

pip install numpy

pip install matplotlib

pip install gpy (or https://github.com/SheffieldML/GPy)

The figures generated by the examples will be saved in the figures folder.

Description of the examples

** GMR01 **

Example of GMR for 2-D outputs with time as input. Corresponds to Figure 1a of the main paper.

** GPR_coregionalization01 **

Example of multi-output GP for 2-D output with time as input. Corresponds to Figure 2b of the main paper.

** GMR_based_GPR01 **

Example of the GMR-based GP for 2-D outputs with time as input. Corresponds to Figure 2 of the main paper.

**GMR_based_GPR_uncertainty_examples01 **

Illustration of GMR-based GPR properties with 1-dimensional input and output. Corresponds to Figure 3 of the main paper. The user can change the number of observations, the lengthscale parameter and the noise variance in the file to generate different examples.