GPmp: the Gaussian Process micro package

GPmp is a lightweight toolkit for Gaussian process (GP) modeling. It provides the essential components for GP-based algorithms with a focus on speed and customization.

Features

• GP interpolation & regression: Supports known or unknown mean functions (including intrinsic kriging).
• Likelihood computation: standard Gaussian and restricted likelihood.
• Efficient cross-validation: fast leave-one-out predictions.
• Conditional sampling: generate conditional sample paths.

It is up to the user to write the mean and covariance functions for setting a GP model.

However, for the purpose of the example, GPmp provides functions for:

• anisotropic scaling
• distance matrix
• Matérn kernels with half-integer regularities
• parameter selection procedure using maximum likelihood, restricted maximum likelihood, or user-defined criteria
• model diagnosis
• vizualization of results
• ...

Backends

GPmp supports three numerical backends:

• PyTorch: Dynamic computation with auto-differentiation.
• JAX: Auto-differentiation with JIT compilation.
• NumPy: Basic computation (default if neither PyTorch nor JAX are found).

On startup, GPmp automatically selects the backend in this order: PyTorch → JAX → NumPy. You can override this by setting the GPMP_BACKEND environment variable before launching GPmp.

Installation

Clone the repository:

git clone https://github.com/gpmp-dev/gpmp.git

Install in development mode:

pip install -e .

Backend Setup

• PyTorch: Recommended for best performance.
• JAX:
  • For CPU-only usage, install directly from PyPI:

    pip install jax

  • For NVIDIA GPU support (e.g., with CUDA 12), run:

    pip install -U "jax[cuda12]"

  Detailed, platform-specific instructions are available in the official JAX documentation.

Usage

Check the examples in the repository for a quick start. Customize your own mean and covariance functions as needed.

Documentation

GPmp’s documentation is built with Sphinx using the PyData theme. To generate the HTML documentation:

cd doc
make html

How to Cite

If you use GPmp in your research, please cite it as follows:

@software{gpmp2025,
  author       = {Emmanuel Vazquez},
  title        = {GPmp: the Gaussian Process micro package},
  year         = {2025},
  url          = {https://github.com/gpmp-dev/gpmp},
  note         = {Version x.y},
}

Please update the version number as appropriate.

Future Work

• Expand documentation and tutorials.
• Enhance diagnostic tools and model visualization.

Authors

See AUTHORS.md for details.

License

GPmp is free software released under the GNU General Public License v3.0. See LICENSE for more details.