/svgp

Code to fit scalable variational Gaussian processes (SVGP)

Primary LanguageJupyter Notebook

Scalable Variational Gaussian Process library (SVGP)

This library contains code to fit Gaussian processes (GPs) using variational inference. The key reference is:

Hensman, James, Alexander Matthews, and Zoubin Ghahramani. "Scalable variational Gaussian process classification." (2015). Available here: http://proceedings.mlr.press/v38/hensman15.pdf

The code provides a simple implementation in Tensorflow 2. We also extend the methodology to do approximate variational inference in hierarchical multi-output Gaussian processes (MOGPs).