VFE approximation for Gaussian processes, the gory details
This post gives the VFE Gaussian process derivation in detail. The implementation details are given in another post. \(\newcommand{\X …
Final Report GSOC 2018
Report of work done so far GSOC 2018 Summary This summer I focused primarily on improvements to the PyMC3 Gaussian …
Mauna Loa Example 2: Incorporating atmospheric measurements
This GP example shows how to: Fit fully Bayesian GPs with NUTS Model inputs which are themselves uncertain (uncertainty in …
Mauna Loa Example 2: Ice core data
This GP example shows how to: Fit fully Bayesian GPs with NUTS Model inputs which are themselves uncertain (uncertainty in …
Looking at the Keeling Curve with GPs in PyMC3
This post discusses modeling the CO2 measurments at Mauna Loa using Gaussian processes in PyMC3.
GP module refactor
An outline of my refactor of the GP module so far.
GPs with non-Normal likelihoods in PyMC3
Gaussian processes can be used with non-Gaussian likelihoods. In this case, the latent variables cannot be marginalized away.
PyMC3 FITC/VFE implementation notes
This post shows in detail how FITC and VFE is implemented in PyMC3.
FITC and VFE
Two general Gaussian Process approximation methods are FITC (fully independent training conditional), and VFE (variational free energy).
Inducing point methods to speed up GPs
Another main avenue for speeding up GPs is inducing point methods, or sparse GPs.
Speeding up GPs with special kernel matrix structure
Linear algebra tricks to speed up GPs
Gaussian processes models
A from-the-ground-up description of Bayesian gaussian process models.