IGP (version 0.2.1)

IGP_hetGP: IGP R6 object for fitting hetGP model

Description

Class providing object with methods for fitting a GP model

Value

Object of R6Class with methods for fitting GP model.

Format

R6Class object.

Methods

Documentation

For full documentation of each method go to https://github.com/CollinErickson/IGP/

new(X=NULL, Z=NULL, package=NULL, estimate.nugget=T, nugget0=F, ...)

This method is used to create object of this class with X and Z as the data. The package tells it which package to fit the GP model.

update(Xall=NULL, Zall=NULL, Xnew=NULL, Znew=NULL, ...)

This method updates the model, adding new data if given, then running optimization again.

Super class

IGP::IGP -> IGP_hetGP

Public fields

X

Design matrix

Z

Responses

N

Number of data points

D

Dimension of data

Active bindings

X

Design matrix

Z

Responses

N

Number of data points

D

Dimension of data

Methods

Public methods

• IGP_hetGP$.init()

• IGP_hetGP$.update()

• IGP_hetGP$.predict()

• IGP_hetGP$.predict.se()

• IGP_hetGP$.predict.var()

• IGP_hetGP$.grad()

• IGP_hetGP$.delete()

• IGP_hetGP$.theta()

• IGP_hetGP$.nugget()

• IGP_hetGP$.s2()

• IGP_hetGP$.mean()

• IGP_hetGP$clone()

Inherited methods

• IGP::IGP$at.max.var()
• IGP::IGP$delete()
• IGP::IGP$finalize()
• IGP::IGP$grad()
• IGP::IGP$grad_from_theta()
• IGP::IGP$grad_norm()
• IGP::IGP$grad_num()
• IGP::IGP$init()
• IGP::IGP$initialize()
• IGP::IGP$max.var()
• IGP::IGP$mean()
• IGP::IGP$nugget()
• IGP::IGP$plot()
• IGP::IGP$predict()
• IGP::IGP$predict.se()
• IGP::IGP$predict.var()
• IGP::IGP$prop.at.max.var()
• IGP::IGP$s2()
• IGP::IGP$sample()
• IGP::IGP$theta()
• IGP::IGP$update()

---

Method .init()

Usage

IGP_hetGP$.init(
  ...,
  kernel = "Gaussian",
  nug.est = self$estimate.nugget,
  nug = self$nugget0,
  noiseControl = list(k_theta_g_bounds = c(1, 100), g_max = 100, g_bounds = c(1e-06, 1)),
  lower = rep(1e-06, ncol(self$X)),
  upper = rep(50, ncol(self$X))
)

---

Method .update()

Usage

IGP_hetGP$.update(...)

---

Method .predict()

Usage

IGP_hetGP$.predict(XX, se.fit, ...)

---

Method .predict.se()

Usage

IGP_hetGP$.predict.se(XX, ...)

---

Method .predict.var()

Usage

IGP_hetGP$.predict.var(XX, ...)

---

Method .grad()

Usage

IGP_hetGP$.grad(XX)

---

Method .delete()

Usage

IGP_hetGP$.delete(...)

---

Method .theta()

Usage

IGP_hetGP$.theta()

---

Method .nugget()

Usage

IGP_hetGP$.nugget()

---

Method .s2()

Usage

IGP_hetGP$.s2()

---

Method .mean()

Usage

IGP_hetGP$.mean()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

IGP_hetGP$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

n <- 40
d <- 2
n2 <- 20
f1 <- function(x) {x[1]*sin(2*pi*x[1]) + sqrt(x[1])*exp(x[2])}
X1 <- matrix(runif(n*d),n,d)
Z1 <- apply(X1,1,f1) + rnorm(n, 0, 1e-2)
X2 <- matrix(runif(n2*d),n2,d)
Z2 <- apply(X2,1,f1)
XX1 <- matrix(runif(10),5,2)
ZZ1 <- apply(XX1, 1, f1)
u <- IGP_hetGP$new(X=X1,Z=Z1)
cbind(u$predict(XX1), ZZ1)
u$predict.se(XX1)
# \donttest{
ContourFunctions::cf(function(x) u$predict(x), pts=X1)
ContourFunctions::cf(function(x) u$predict(x, se.fit=TRUE)$se, pts=X1)
# }
u$update(Xnew=X2,Znew=Z2)
u$predict(XX1)
ContourFunctions::cf(function(x) u$predict(x), pts=rbind(X1, X2))
u$delete()

n <- 10
d <- 1
X1 <- runif(n)
X1 <- c(X1, X1)
f1 <- function(x) {abs(sin(pi*x))+sqrt(x)+rnorm(1,0,.1)}
Z1 <- sapply(X1, f1)
plot(X1, Z1)
h1 <- hetGP::mleHetGP(X=matrix(X1,ncol=1),Z=matrix(Z1, ncol=1), lower=c(.1), upper=c(50))
curve(predict(h1, matrix(x, ncol=1))$mean, add=TRUE, col=3)
u <- IGP_hetGP$new(X=X1,Z=Z1)
plot(X1, Z1, col=2); curve(u$predict(matrix(x,ncol=1)), add=TRUE)