vanguard-gp 3.0.0

Various easy-to-use extensions for Gaussian process models and a framework for composition of extensions.

Vanguard: Advanced GPs

Vanguard is a high-level wrapper around GPyTorch and aims to provide a user-friendly interface for training and using Gaussian process models. Vanguard's main objective is to make a variety of more advanced GP techniques in the machine learning literature available for easy use by a non-specialists and specialists alike. Vanguard is designed for modularity to facilitate straightforward combinations of different techniques.

Vanguard implements many advanced Gaussian process techniques, as showcased in our examples folder. These techniques and others implemented within the Vanguard paradigm can be combined straightforwardly with minimal extra code, and without requiring specialist GP knowledge.

Installation

To install Vanguard:

$ pip install vanguard-gp

Note that it is vanguard-gp and not vanguard. However, to import the package, use from vanguard import ....

There are optional sets of additional dependencies:

• vanguard-gp[test] is required to run the tests;
• vanguard-gp[doc] is for compiling the Sphinx documentation;
• vanguard-gp[notebook] contains all dependencies for the example notebooks;
• vanguard-gp[dev] includes all tools and packages a developer of Vanguard might need.

Should the installation of Vanguard fail, you can see the versions used by the Vanguard development team in uv.lock. You can transfer these to your own project as follows. First, install UV. Then, clone the repo from GitHub. Next, run

uv export --format requirements-txt

which will generate a requirements.txt. Install this in your own project before trying to install Vanguard itself,

pip install -r requirements.txt
pip install vanguard-gp

Documentation

Vanguard's documentation can be found online.

Alternatively, you can build the documentation from source - instructions for doing so can be found in CONTRIBUTING.md.

Examples

Vanguard contains a number of example notebooks, contained in the examples/notebooks folder. These are designed to showcase certain features of Vanguard within the context of a data science problem. To run them, you will need to first install the additional requirements:

$ pip install .[notebook]

If you are in a virtual environment, you can then run the following to add the vanguard kernel to Jupyter, which makes running the notebooks as frictionless as possible:

$ ipython kernel install --name vanguard --user

Warning: Certain notebooks can take a long time to run, even on a GPU. To see fully rendered examples, please visit the documentation.