optims: Title Multi-local optimization wrapper for optim, using...

In IRSN/DiceView: Methods for Visualization of Computer Experiments Design and Surrogate

Title Multi-local optimization wrapper for optim, using (possibly parallel) multistart.

Description

Title Multi-local optimization wrapper for optim, using (possibly parallel) multistart.

Usage

optims(
  pars,
  fn,
  fn.NaN = NaN,
  .apply = "mclapply",
  pars.eps = 1e-05,
  control = list(),
  ...
)

Arguments

pars	starting points for optim
fn	objective function, like in optim().
fn.NaN	replacement value of fn when returns NaN
.apply	loop/parallelization backend for multistart ("mclapply", "lapply" or "foreach")
pars.eps	minimal distance between two solutions to be considered different
control	control parameters for optim()
...	additional arguments passed to optim()

Value

list with best solution and all solutions

Author(s)

Yann Richet, IRSN

Examples

fn = function(x) ifelse(x==0,1,sin(x)/x)
# plot(fn, xlim=c(-20,20))
optim( par=5,                     fn,lower=-20,upper=20,method='L-BFGS-B')
optims(pars=t(t(seq(-20,20,,20))),fn,lower=-20,upper=20,method='L-BFGS-B')

# Branin function (3 local minimas)
f = function (x) {
  x1 <- x[1] * 15 - 5
  x2 <- x[2] * 15
  (x2 - 5/(4 * pi^2) * (x1^2) + 5/pi * x1 - 6)^2 + 10 * (1 - 1/(8 * pi)) * cos(x1) + 10
}
# expect to find 3 local minimas
optims(pars=matrix(runif(100),ncol=2),f,method="L-BFGS-B",lower=c(0,0),upper=c(1,1))

IRSN/DiceView documentation built on Jan. 31, 2024, 10:09 a.m.