[R] any more direct-search optimization method in R

[Ben Bolker]

Ingmar Visser <I.Visser <at> uva.nl> writes:

> 
> If you have only boundary constraints on parameters you can use method
> L-BFGS in optim.
> Hth, ingmar
> 
> > From: Weijie Cai <wcai11 <at> hotmail.com>

> > 
> > I am dealing with a noisy function (gradient,hessian not available) with
> > simple boundary constraints (x_i>0). I've tried constrOptim() using nelder
> > mead to minimize it but it is way too slow and the returned results are not
> > satisfying. simulated annealing is so hard to tune and it always crashes R
> > program in my case. I wonder if there are any packages or functions can do
> > direct search optimization?
> > 

   Noisy functions are really challenging to optimize; (1) there is no
"best" method (despite all the papers doing comparisons of stochastic
global optimizers on various sets of test functions); (2) the fancier
methods are hard to program [and existing implementations tend have more
restricted licenses]; (3) they tend to be slow (thousands of function
evaluations).   Packages on CRAN that *might* be helpful are
genalg, DEoptim.
   A typical "poor man's" approach to boundary constraints is to
add a quadratic penalty (perhaps not even trying to evaluate the
objective function -- e.g. substituting the value at the closest
boundary point) for parameters outside the constraints into
the objective function.
   With more information (number of parameters, time to compute a
single function evaluation, kind of noise) we might be able to help
more.
  
  Ben Bolker