SHGP - Sparse Hierarchical Gaussian Process

A sparse implementation of a heirarchical Gaussian processes in gpflow. On the jax branch we are planning to produce a jax implementation.

Please note that there is an active issue where the compilation of models with a large number of time series (realisations) create memory problems. We're working on this

Setup

Create conda environment

conda env create -f environment.yml
conda activate shgp-env

Install package:

python setup.py install .

Examples

• This is a short (fully interactive) example based on a talk.
• There is alsoa basic data example at examples/basic_example.ipynb and a climate modelling based example at examples/climate_modelling_example_1D.ipynb