Hi all, I need to minimize following function :
dat <- matrix(rnorm(20000), ncol=2)
targetFn <- function(x) {
dat <- as.matrix(dat)
dat1 <- 1*dat[,1] - (x^2)*dat[,2]
return(sd(dat1)) }
i.e. I want ro find for which "x" the value of "targetFn"
will be minimum,
depending on current dataset "dat". Is there any optimization routine
available for this type of optimization?
Thanks and regards,
[[alternative HTML version deleted]]
Bogaso wrote:> > Hi all, I need to minimize following function : > > dat <- matrix(rnorm(20000), ncol=2) > targetFn <- function(x) { > dat <- as.matrix(dat) > dat1 <- 1*dat[,1] - (x^2)*dat[,2] > return(sd(dat1)) } > > i.e. I want ro find for which "x" the value of "targetFn" will be minimum, > depending on current dataset "dat". Is there any optimization routine > available for this type of optimization? >Try the help ??optimization And some mathematics will show that in your particular case x=0 is the optimal value. Berend -- View this message in context: http://r.789695.n4.nabble.com/Optimization-tp2228300p2228350.html Sent from the R help mailing list archive at Nabble.com.
Berend Hasselman wrote:> > > Bogaso wrote: >> >> Hi all, I need to minimize following function : >> >> dat <- matrix(rnorm(20000), ncol=2) >> targetFn <- function(x) { >> dat <- as.matrix(dat) >> dat1 <- 1*dat[,1] - (x^2)*dat[,2] >> return(sd(dat1)) } >> >> i.e. I want ro find for which "x" the value of "targetFn" will be >> minimum, >> depending on current dataset "dat". Is there any optimization routine >> available for this type of optimization? >> > ... > > And some mathematics will show that in your particular case x=0 is the > optimal value. >Correction: x=0 is one of the optimal values. Whether the other alternative for x is feasible, depends on the sign of the crossproduct of dat[,1] and dat[,2]. If the rnorm() behaves as is to be expected that crossproduct will be close to zero. Berend -- View this message in context: http://r.789695.n4.nabble.com/Optimization-tp2228300p2228600.html Sent from the R help mailing list archive at Nabble.com.