The R Project SVN R

\name{optimize}
\title{One Dimensional Optimization}
\usage{
optimize(f = , interval = , lower = min(interval),
        upper = max(interval), maximum = FALSE,
        tol = .Machine$double.eps^0.25, \dots)
optimise(f = , interval = , lower = min(interval),
        upper = max(interval), maximum = FALSE,
        tol = .Machine$double.eps^0.25, \dots)
}
\alias{optimize}
\alias{optimise}
\arguments{
  \item{f}{the function to be optimized. The function is
    either minimized or maximized over its first argument
    depending on the value of \code{maximum}.}
  \item{interval}{a vector containing the end-points of the interval
    to be searched for the minimum.}
  \item{lower}{the lower end point of the interval
    to be searched.}
  \item{upper}{the upper end point of the interval
    to be searched.}
  \item{maximum}{logical. Should we maximize or minimize (the default)?}
  \item{tol}{the desired accuracy.}
  \item{\dots}{additional arguments to \code{f}.}
}
\description{
  The function \code{optimize} searches the interval from
  \code{lower} to \code{upper} for a minimum or maximum of
  the function \code{f} with respect to its first argument.

  \code{optimise} is an alias for \code{optimize}.
}
\details{
  The method used is a combination of golden section search and
  successive parabolic interpolation.  Convergence is never much slower
  than that for a Fibonacci search.  If \code{f} has a continuous second
  derivative which is positive at the minimum (which is not at \code{lower} or
  \code{upper}), then convergence is superlinear, and usually of the
  order of about 1.324.

  The function \code{f} is never evaluated at two points closer together
  than \eqn{\epsilon}{eps *}\eqn{ |x_0| + (tol/3)}, where
  \eqn{\epsilon}{eps} is approximately \code{sqrt(\link{.Machine}$double.eps)}
  and \eqn{x_0} is the final abscissa \code{optimize()$minimum}.\cr
  If \code{f} is a unimodal function and the computed values of \code{f}
  are always unimodal when separated by at least \eqn{\epsilon}{eps *}
  \eqn{ |x| + (tol/3)}, then \eqn{x_0} approximates the abcissa of the
  global minimum of \code{f} on the interval \code{lower,upper} with an
  error less than \eqn{\epsilon}{eps *}\eqn{ |x_0|+ tol}.\cr
  If \code{f} is not unimodal, then \code{optimize()} may approximate a
  local, but perhaps non-global, minimum to the same accuracy.

  The first evaluation of \code{f} is always at
  \eqn{x_1 = a + (1-\phi)(b-a)} where \code{(a,b) = (lower, upper)} and
  \eqn{\phi = (\sqrt 5 - 1)/2 = 0.61803..} is the golden section ratio.
  Almost always, the second evaluation is at \eqn{x_2 = a + \phi(b-a)}.
  Note that a local minimum inside \eqn{[x_1,x_2]} will be found as
  solution, even when \code{f} is constant in there, see the last
  example.

  It uses a C translation of Fortran code (from Netlib) based on the
  Algol 60 procedure \code{localmin} given in the reference.
}
\value{
  A list with components \code{minimum} (or \code{maximum})
  and \code{objective} which give the location of the minimum (or maximum)
  and the value of the function at that point.
}
\references{
  Brent, R. (1973)
  \emph{Algorithms for Minimization without Derivatives.}
  Englewood Cliffs N.J.: Prentice-Hall.
}
\seealso{
  \code{\link{nlm}}, \code{\link{uniroot}}.
}
\examples{
f <- function (x,a) (x-a)^2
xmin <- optimize(f, c(0, 1), tol = 0.0001, a = 1/3)
xmin

## See where the function is evaluated:
optimize(function(x) x^2*(print(x)-1), l=0, u=10)

## "wrong" solution with unlucky interval and piecewise constant f():
f  <- function(x) ifelse(x > -1, ifelse(x < 4, exp(-1/abs(x - 1)), 10), 10)
fp <- function(x) { print(x); f(x) }

plot(f, -2,5, ylim = 0:1, col = 2)
optimize(fp, c(-4, 20))# doesn't see the minimum
optimize(fp, c(-7, 20))# ok
}
\keyword{optimize}