ROI optimisation in R using multi-argument F_objective function - r

Question

Trying to run a simple ROI optimisation in R, but after hours of fidgeting I'm at a loss. I keep getting the error:
Error in .check_function_for_sanity(F, n) :
cannot evaluate function 'F' using 'n' = 5 parameters.
Here is the sample code:
library(ROI)
library(nloptr)
library(ROI.plugin.nloptr)
#Generate some random data for this example
set.seed(3142)
myRet = matrix(runif(100 * 5, -0.1, 0.1), ncol = 5)
myCovMatrix = cov(myRet)
myRet <- myRet
myCovMatrix <- myCovMatrix
# Sample weights
w <- rep(1/ncol(myRet), ncol(myRet))
#Define functions for the optimisation
diversificationRatio = function(w, covMatrix)
{
weightedAvgVol = sum(w * sqrt(diag(covMatrix)))
portfolioVariance = (w %*% covMatrix %*% w)[1,1]
- 1 * weightedAvgVol / sqrt(portfolioVariance)
}
# Check that the F_objective function works:
diversificationRatio(w, myCovMatrix)
# Now construct the F_objective
foo <- F_objective(F = diversificationRatio, n = (ncol(myRet)))
Any ideas on how many parameters to pass to n?

Answer 1

F_objective expects a function with only one argument so you have to write a wrapper function.
#Define functions for the optimisation
diversificationRatio <- function(w, covMatrix) {
weightedAvgVol <- sum(w * sqrt(diag(covMatrix)))
portfolioVariance <- (w %*% covMatrix %*% w)[1,1]
- 1 * weightedAvgVol / sqrt(portfolioVariance)
}
# Check that the F_objective function works:
wrapper <- function(x) diversificationRatio(x, myCovMatrix)
# Now construct the F_objective
o <- OP(F_objective(F = wrapper, n = (ncol(myRet))))
ROI_applicable_solvers(o)
start <- runif(ncol(myRet))
s <- ROI_solve(o, solver = "nloptr", start = start, method = "NLOPT_LD_SLSQP")
s
solution(s)