GP Playground

This project is an educational resource which explores selected aspects of Gaussian processes (GPs). It also provides teaching material which has been used in several iterations of this online course.

Contents

• The role of noise in GPs
  • Reference implementation of RW06 equations
  • Comparison of GP libraries
  • Prior and (noisy) posterior predictions

• Compare GPs and kernel ridge regression (KRR)

• Teaching material
  • 1D exact GP regression
  • 2D exact GP regression
  • 1D approximate GP regression using Sparse Stochastic Variational Inference

Resources

• Our own

  • Intro to GPs and kernel methods talk

  • GP talk

  • Uncertainty Quantification for Neural Networks (with GP intro)

• Blog post about GP vs. KRR

• This section was inspired by a discussion over at the sklearn issue tracker. Thanks!

• Books and publications listed below: [Bis06, DLvdW20, GKD19, KHSS18, Mur23, RW06]

[Bis06]

Christopher Bishop. Pattern Recognition and Machine Learning. Springer, 2006. URL: https://www.microsoft.com/en-us/research/publication/pattern-recognition-machine-learning/.

[DLvdW20]

Marc Peter Deisenroth, Yicheng Luo, and Mark van der Wilk. A practical guide to gaussian processes. 2020. URL: https://infallible-thompson-49de36.netlify.app/ (visited on 2020-12-03).

[GKD19]

Jochen Görtler, Rebecca Kehlbeck, and Oliver Deussen. A Visual Exploration of Gaussian Processes. 2019. URL: https://distill.pub/2019/visual-exploration-gaussian-processes (visited on 2022-11-22), doi:10.23915/distill.00017.

[KHSS18]

Motonobu Kanagawa, Philipp Hennig, Dino Sejdinovic, and Bharath K. Sriperumbudur. Gaussian Processes and Kernel Methods: A Review on Connections and Equivalences. 2018. URL: http://arxiv.org/abs/1807.02582, arXiv:1807.02582.

[Mur23]

Kevin P. Murphy. Probabilistic Machine Learning: Advanced Topics. MIT Press, 2023. URL: https://probml.github.io/pml-book/book2.html.

[RW06]

Carl Edward Rasmussen and Christopher K. I. Williams. Gaussian Processes for Machine Learning. MIT Press, 2006. ISBN ISBN-10 0-262-18253-X. URL: http://www.gaussianprocess.org/gpml.

Citing

The source code and this book are licensed under the BSD 3-Clause License. If you re-use material from this work or just like to cite it, then either use this BibTeX entry

@Online{schmerler_GPPlayground,
  author   = {Steve Schmerler},
  title    = {GP Playground},
  url      = {https://github.com/elcorto/gp_playground},
  subtitle = {Explore selected topics related to Gaussian processes},
  doi      = {10.5281/zenodo.7439202},
}

or the DOI 10.5281/zenodo.7439202.