This is an example of the type of problem, and how i'm currently using
optim() to solve it.
mydata <- runif(500,-1,1)
myfunc <- function(p,d)
{
print(p <- floor(p))
ws <- function(i,n,x) sum(x[i-n+1]:x[i])
ws1 <- c(rep(NA,p[1]-1),sapply(p[1]:NROW(d),ws,p[1],d))
ws2 <- c(rep(NA,p[2]-1),sapply(p[2]:NROW(d),ws,p[2],d))
ws3 <- c(rep(NA,p[3]-1),sapply(p[3]:NROW(d),ws,p[3],d))
var(ws1+ws2+ws3,na.rm=TRUE)
}
opt <- optim(c(25,50,150),myfunc,method="L-BFGS-B",
control=list(fnscale=-1,parscale=c(1,1,1),factr=1,ndeps=c(5,5,5)),
lower=t(c(1,51,101)),upper=t(c(50,100,200)),d=mydata)
print(floor(opt$par))
print(myfunc(opt$par,mydata))
So the parameters to the function to be optimized are parameters to
functions that only accept integer values. All of the paramters to be
optimized are integers that are subject to upper lower bound constraints.
This was the solution I came up with after checking CRAN, searching nabble
etc.
It runs but not very efficiently, and does not seem to really explore the
sample space very well before focusing on a local minimum. I also looked at
the constrOptim function but I couldn't figure out how to implement two
sided constraints with it.
--
View this message in context:
http://r.789695.n4.nabble.com/non-linear-integer-optimization-tp2551409p2552080.html
Sent from the R help mailing list archive at Nabble.com.
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
