fnchk: Run tests, where possible, on user objective function

In optimx: Expanded Replacement and Extension of the 'optim' Function

Run tests, where possible, on user objective function

Description

fnchk checks a user-provided R function, ffn.

Usage

   fnchk(xpar, ffn, trace=0, ... )

Arguments

xpar	the (double) vector of parameters to the objective funcion
ffn	a user-provided function to compute the objective function
trace	set >0 to provide output from fnchk to the console, 0 otherwise
...	optional arguments passed to the objective function.

Details

fnchk attempts to discover various errors in function setup in user-supplied functions primarily intended for use in optimization calculations. There are always more conditions that could be tested!

Value

The output is a list consisting of list(fval=fval, infeasible=infeasible, excode=excode, msg=msg)

fval	The calculated value of the function at parameters xpar if the function can be evaluated.
infeasible	FALSE if the function can be evaluated, TRUE if not.
excode	An exit code, which has a relationship to
msg	A text string giving information about the result of the function check: Messages and the corresponding values of excode are: fnchk OK; excode = 0; infeasible = FALSE Function returns INADMISSIBLE; excode = -1; infeasible = TRUE Function returns a vector not a scalar; excode = -4; infeasible = TRUE Function returns a list not a scalar; excode = -4; infeasible = TRUE Function returns a matrix list not a scalar; excode = -4; infeasible = TRUE Function returns an array not a scalar; excode = -4; infeasible = TRUE Function returned not length 1, despite not vector, matrix or array; excode = -4; infeasible = TRUE Function returned non-numeric value; excode = 0; excode = -1; infeasible = TRUE Function returned Inf or NA (non-computable); excode = -1; infeasible = TRUE

Author(s)

John C. Nash <[e-mail]>

Examples

# Want to illustrate each case.
# Ben Bolker idea for a function that is NOT scalar
# rm(list=ls())
# library(optimx)
sessionInfo()
benbad<-function(x, y){
  # y may be provided with different structures
  f<-(x-y)^2
} # very simple, but ...

y<-1:10
x<-c(1)
cat("fc01: test benbad() with y=1:10, x=c(1)\n")
fc01<-fnchk(x, benbad, trace=4, y)
print(fc01)

y<-as.vector(y)
cat("fc02: test benbad() with y=as.vector(1:10), x=c(1)\n")
fc02<-fnchk(x, benbad, trace=1, y)
print(fc02)

y<-as.matrix(y)
cat("fc03: test benbad() with y=as.matrix(1:10), x=c(1)\n")
fc03<-fnchk(x, benbad, trace=1, y)
print(fc03)

y<-as.array(y)
cat("fc04: test benbad() with y=as.array(1:10), x=c(1)\n")
fc04<-fnchk(x, benbad, trace=1, y)
print(fc04)

y<-"This is a string"
cat("test benbad() with y a string, x=c(1)\n")
fc05<-fnchk(x, benbad, trace=1, y)
print(fc05)

cat("fnchk with Rosenbrock\n")
fr <- function(x) {   ## Rosenbrock Banana function
  x1 <- x[1]
  x2 <- x[2]
  100 * (x2 - x1 * x1)^2 + (1 - x1)^2
}
xtrad<-c(-1.2,1)
ros1<-fnchk(xtrad, fr, trace=1)
print(ros1)
npar<-2
opros<-list2env(list(fn=fr, gr=NULL, hess=NULL, MAXIMIZE=FALSE, PARSCALE=rep(1,npar), FNSCALE=1,
                     KFN=0, KGR=0, KHESS=0, dots=NULL))
uros1<-fnchk(xtrad, fr, trace=1)
print(uros1)

optimx documentation built on Oct. 24, 2023, 5:06 p.m.