minimization of a set of functions r

Question

I need to find an efficient way to find the minimum of a set of homogeneous functions.

I am using the function stats::optimize

If I do it with one single function, no problem. However I need to change function parameters taking them from a dataframe and I can't find a way to let optimize process my set of functions. See code below

myfun <- function (x) {
  sqrt((x-3)^2 + (x-4)^2 + (x-5)^2)
}

optimize(myfun, c(0,10)

no problems here.

However I need to substitute those numbers with each row of a data frame such as

df <- data.frame(matrix(c(2,7,8,4,9,10,5,4,2), nrow = 3, ncol =3))

something like:

v <-df[1, ]
myfun2 <- function (v) {
  function(x) sqrt((x-v[1])^2 + (x-v[2])^2 + (x-v[3])^2)
}

optimize(myfun2, c(0,10))

Error in optimize(myfun2, c(0, 10)) : 
  invalid function value in 'optimize'

optimize(myfun2(df[1, ]), c(0,10))
Error in optimize(myfun2(df[1, ]), c(0, 10)) : 
  invalid function value in 'optimize'

for a single case, that would eventually end up in a for loop for covering each row of the data frame

however optimize returns an error if I pass myfun2.

Sorry if this is a simple question, but I really cannot find the right way to solve it and any help would be very much appreciated.

I also I tried

m <- matrix(c(2,7,8,4,9,10,5,4,2), nrow = 3, ncol =3)
myfun2 <- function (v) {
  function(x) sqrt((x-m[1,1])^2 + (x-m[1,2])^2 + (x-m[1,3])^2)
}       
optimize(myfun2, c(0,10))

Error in optimize(myfun2, c(0, 10)) : 
  invalid function value in 'optimize'

Provide the error and perhaps the data in `v`... make it reproducible. — cory, Jun 13 '19 at 15:13
(a) Can you please edit your `df <- data.frame...` line so it runs? Not valid syntax right now. (b) "optimize returns an error if I pass myfun2", what is the error? (c) What happens if you run `myfun2(df[1, ])`, does it work, or do you get an error? — Gregor Thomas, Jun 13 '19 at 15:13
My suspicion is you have a problem because `v[1]` is a 1-row 1-column data frame. If you switch `v` to a `matrix` it may work... — Gregor Thomas, Jun 13 '19 at 15:16
Thanks for all the suggestions and apologies for the typos. I edited the question, but nothing worked so far. Any other suggestion? — CZuC, Jun 13 '19 at 16:14
Your several attempts of `myfun2` return a *function*, not a *number*. — Rui Barradas, Jun 13 '19 at 16:27
could you please provide an example that explains why I am wrong? — CZuC, Jun 13 '19 at 16:37

score 1 · Accepted Answer · answered Jun 13 '19 at 16:32

The function that has to be used is the original myfun, with the numbers replaced by v[1], v[2] and v[3], called in an apply loop.

myfun <- function (x, v) {
  sqrt((x - v[1])^2 + (x - v[2])^2 + (x - v[3])^2)
}

df <- data.frame(matrix(c(2,7,8,4,9,10,5,4,2), nrow = 3, ncol =3))

res <- apply(df, 1, function(.v) optimize(myfun, c(0,10), v = .v))

do.call(rbind, res)
#     minimum  objective
#[1,] 3.666648 2.160247 
#[2,] 6.666666 3.559026 
#[3,] 6.666667 5.887841

minimization of a set of functions r

1 Answers1