Universal Prediction distribution based Efficient Global Optimization UP-EGO

Global optimization using the Universal Prediction distribution

upego(model, fun, nsteps, lower, upper, seed = 1, parinit = NULL,
  control = NULL, EEIcontrol = NULL)

Arguments

model	the surrogate model
fun	the real function
nsteps	the number of points to be generated
lower	the lower bound of the design space
upper	the upper bound of the design space
seed	the random seed (default = 1)
parinit	inital points to be used in the optimization (default NULL)
control	the optimization control parameters (default NULL)
EEIcontrol	the optimization criterion parameters (default NULL)

Value

list of generated points and their values and the last updated surrogate model

Examples

#' library(UP)
d            <- 2;
n            <- 16
X            <- expand.grid(x1=s <- seq(0,1, length=5), x2=s)
Y            <- apply(X, 1, branin)

upsm            <- UPSM$new(sm= krigingsm$new(), UP=UPClass$new(X,Y,Scale =TRUE) )
upego_res <-  upego( upsm,fun = branin, nsteps = 1, lower= c(0,0),upper = c(1,1) )
#> [1] "OPTIMIZATION step : 1 ..."
#> [1] "new point "
#>         x1    x2
#> [1,] 0.125 0.625
#> [1] "New value  8.43883174131785"
#> [1] "current best value :  2.94383349064161   inital :  2.94383349064161"

print( min(upego_res$last$get_DOE()$y ) )
#> [1] 2.943833