problem with nloptr package in a maximization problem

[Llorenxo]

Good morning to everyone,
I've a problem with a maximization with the R package nloptr. I've to maximize a correlation between a variable, call it "a", and a linear combination of other variables. Changing the weigths of the varibles in order to maximize the correlation. This is an example:

library(nloptr)
#create a dataset for the example

data=data.frame("a"=c(1:10), "b"=c(2,3,4,2,3,1,2,4,1,6), "c"=rep(c(10,15), 5))

# Objective Function
eval_f <- function(x,y)
{
  return (cor(data$a,(x*data$b+y*data$c)))
}

eval_f(2,2)
# Equality constraints
eval_g_eq <- function(x,y)
{
  return ( x+y-1 )
}
# Lower and upper bounds
lb <- c(0,0)
ub <- c(1,1)
#initial values
x0 <- c(0.5,0.5)

# Set optimization options.
local_opts <- list( "algorithm" = "NLOPT_LD_MMA", "xtol_rel" = 1.0e-15 )
opts <- list( "algorithm"= "NLOPT_GN_ISRES",
              "xtol_rel"= 1.0e-15,
              "maxeval"= 160000,
              "local_opts" = local_opts,
              "print_level" = 0 )
res <- nloptr ( x0 = x0,
                eval_f = eval_f,
                lb = lb,
                ub = ub,
                eval_g_eq = eval_g_eq,
                opts = opts
)

This give me the error:

Error in .checkfunargs(eval_f, arglist, "eval_f") : 
  eval_f requires argument 'y' but this has not been passed to the 'nloptr' function.

Could someone help me?

Thanks.

[nirgrahamuk]

nloprt wants your function to receive params in a certain way, that is in a single variable, where each entry of the variable is a param.
You can do this quite straightforwardly for your case.

# Objective Function
eval_f <- function(input)
{
  x <-input[1]
  y <-input[2]
  
  return (cor(data$a,(x*data$b+y*data$c)))
}

eval_f(c(2,2))
# Equality constraints
eval_g_eq <- function(input)
{  
x <-input[1]
y <-input[2]
  return ( x+y-1 )
}