gplite-package: The 'gplite' package.

In gplite: General Purpose Gaussian Process Modelling

The 'gplite' package.

Description

gplite implements some of the most common Gaussian process (GP) models. The package offers tools for integrating out the latent values analytically using Laplace or expectation propagation (EP) approximation and for estimating the hyperparameters based on maximizing the (approximate) marginal likelihood or posterior. The package also implements some common sparse approximations for larger datasets.

Functions

Here's a list of the most important functions:

gp_init

Set up the GP model.

cf, lik, method, approx

Choose the covariance functions, likelihood (observation model), type of the GP (full or some sparse approximation) and the latent function approximation method (Laplace, EP).

gp_optim, gp_fit

Optimize the model hyperparameters, or just fit the model with the current hyperparameter values.

gp_pred, gp_draw

Make predictions with the fitted model. Can also be used before fitting to obtain prior predictive distribution or draws.

gp_loo, gp_compare

Model assessment and comparison using leave-one-out (LOO) cross-validation.

gplite documentation built on Aug. 24, 2022, 9:07 a.m.