simulate.cobbsGPR: Gaussian Process Simulation

In martinzaefferer/COBBS: Continuous Optimization Benchmarks By Simulation

Gaussian Process Simulation

Description

(Conditional) Simulation at given locations, with a model fit resulting from gaussianProcessRegression. In contrast to prediction or estimation, the goal is to reproduce the covariance structure, rather than the data itself. Note, that the conditional simulation also reproduces the training data, but has a two times larger error than the GPR predictor.

Usage

## S3 method for class 'cobbsGPR'
simulate(
  object,
  nsim = 1,
  seed = NA,
  xsim,
  method = "decompose",
  conditionalSimulation = TRUE,
  Ncos = 10,
  returnAll = FALSE,
  ...
)

Arguments

object	fit of the GPR model (settings and parameters), of class cobbsGPR.
nsim	number of simulations
seed	random number generator seed. Defaults to NA, in which case no seed is set
xsim	list of samples in input space, to be simulated at
method	"decompose" (default) or "spectral", specifying the method used for simulation. Note that "decompose" is can be preferable, since it is exact but may be computationally infeasible for high-dimensional xsim. On the other hand, "spectral" yields a function that can be evaluated at arbitrary sample locations.
conditionalSimulation	logical, if set to TRUE (default), the simulation is conditioned with the training data of the GPR model. Else, the simulation is non-conditional.
Ncos	number of cosine functions (used with method="spectral" only)
returnAll	if set to TRUE, a list with the simulated values (y) and the corresponding covariance matrix (covar) of the simulated samples is returned.
...	further arguments, not used

Value

Returned value depends on the setting of object$simulationReturnAll

References

N. A. Cressie. Statistics for Spatial Data. JOHN WILEY & SONS INC, 1993.

C. Lantuejoul. Geostatistical Simulation - Models and Algorithms. Springer-Verlag Berlin Heidelberg, 2002.

See Also

gaussianProcessRegression, predict.cobbsGPR

martinzaefferer/COBBS documentation built on July 19, 2023, 4:12 a.m.