General

Overview

treegp is a python gaussian process code that perform 1D and 2D interpolation.

treegp have some specific features compared to other gaussian processes available code:

  • Hyperparameters estimation will scale in \({\cal{O}}(N \log(N))\) with the the 2-points correlation function estimation compared to \({\cal{O}}(N^3)\) with the classical maximum likelihood.

  • Gaussian process interpolation can be performed around a mean function

  • Tools are given to compute the mean function (meanify)

treegp was originally developed for Point Spread Function interpolation within Piff. There is a specific article that describes the math used in treegp in the context of modelling astrometric shifts of the Subaru Telescope due to atmospheric turbulences. This article can be found here.

Installation

The easiest way to install is usually:

pip install treegp

which will install the latest released version.

If you would instead like to install the development version, you can do so via:

git clone https://github.com/PFLeget/treegp.git
cd treegp/
python setup.py install

Dependencies

treegp has for now the following dependencies (see the quick installs below):

Python

treegp is regularly tested on Python 3.7, 3.8, 3.9 and 3.10. It may work in other versions of Python (e.g. pypy), but these are not currently supported.