An R package for fitting some of the most common Gaussian process (GP) models. Implements Laplace and EP approximations for handling non-Gaussian observation models, performs hyperparameter optimization using maximum marginal likelihood (or posterior), and implements some common sparse approximations for handling larger datasets. Provides also tools for model assessment and comparison via leave-one-out (LOO) cross-validation.
The syntax has taken a lot of inspiration from that of GPstuff but the intention of the package is not to be a GPstuff clone for R.
Resourcesinstall.packages('gplite')if (!require(devtools)) {
install.packages("devtools")
library(devtools)
}
devtools::install_github('jpiironen/gplite', build_vignettes = TRUE)library(gplite)
library(ggplot2)
# create some toy 1d regression data
set.seed(32004)
n <- 200
sigma <- 0.1
x <- rnorm(n)
y <- sin(3*x)*exp(-abs(x)) + rnorm(n)*sigma
# set up the gp model, and optimize the hyperparameters
gp <- gp_init(cfs = cf_sexp(), lik = lik_gaussian())
gp <- gp_optim(gp, x, y)
# compute the predictive mean and variance in a grid of points
xt <- seq(-4, 4, len=300)
pred <- gp_pred(gp, xt, var=T)
# visualize
mu <- pred$mean
lb <- pred$mean - 2*sqrt(pred$var)
ub <- pred$mean + 2*sqrt(pred$var)
ggplot() +
geom_ribbon(aes(x=xt, ymin=lb, ymax=ub), fill='lightgray') +
geom_line(aes(x=xt, y=mu), size=1) +
geom_point(aes(x=x, y=y), size=0.5) +
xlab('x') + ylab('y')If you find the software useful, please use the following citation:
Piironen, Juho (2021). gplite: General Purpose Gaussian Process Modelling. R package.
Bibtex:
@misc{gplite,
author = {Piironen, Juho},
title = {gplite: General Purpose {G}aussian Process Modelling},
note = {R package},
year = {2021},
url = {https://github.com/jpiironen/gplite},
}Rasmussen, C. E. and Williams, C. K. I. (2006). Gaussian processes for machine learning. MIT Press. Online
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4