R help - Nov 2010 - Fast Two-Dimensional Optimization

Dear R Helpers,

I have attempted "optim" function to solve a two-dimensional
optimization
problem. It took around 25 second to complete the procedure.
However, I want to reduce the computation time: less than 7 second. Is there
any optimization function in R which is very rapid?

Best Regards,
Wonsang


-----
Wonsang You
Leibniz Institute for Neurobiology
-- 
View this message in context:
http://r.789695.n4.nabble.com/R-Fast-Two-Dimensional-Optimization-tp3053782p3053782.html
Sent from the R help mailing list archive at Nabble.com.

	[[alternative HTML version deleted]]

Wonsang You <you <at> ifn-magdeburg.de> writes:
> I have attempted "optim" function to solve a two-dimensional
optimization
> problem. It took around 25 second to complete the procedure.
> However, I want to reduce the computation time: less than 7 second. Is
there
> any optimization function in R which is very rapid?
  This is not nearly enough information for us to help. The answer
depends on the characteristics of your objective function.  You may be
able to quadruple the speed of your optimization by coding your objective
function more efficiently in R or by re-coding it in C or C++. You may
be able to choose better starting conditions.  You may be able to pick
an optimization method that is more suitable for your objective function
(see ?optim and the "optimx" package on r-forge).

  Ben Bolker

Thank you for your kind comments.
I tried different algorithms such as BFGS, CG, and so on. However, the
choice of algorithms was not effective on reducing computation time.

Instead, your second suggestion of coding the gradient of minimization
function was a little bit successful to reduce computation time. It was
twice faster than before. I appreciate you for your help.


2010/11/22 Rubén Roa <rroa@azti.es>
> Did you try different algorithms? optim has Nelder-Mead (default), BFGS,
> simulated annealing, CG, etc.
> Depending on your problem, they can be much faster than the default.
> See
> ?optim
> and check the info for 'method'
>
> Check also optimx, a new wrapper for optim that can try out all methods.
>
> Another approach that may speed up calculations is to code for the
> gradients of the function that you're minimizing.
> Check argument 'gr' of optimx.
>
> HTH
>
>
>
____________________________________________________________________________________
>
> Dr. Rubén Roa-Ureta
> AZTI - Tecnalia / Marine Research Unit
> Txatxarramendi Ugartea z/g
> 48395 Sukarrieta (Bizkaia)
> SPAIN
>
>
>
> > -----Mensaje original-----
> > De: r-help-bounces@r-project.org
> > [mailto:r-help-bounces@r-project.org] En nombre de Wonsang You
> > Enviado el: lunes, 22 de noviembre de 2010 16:17
> > Para: r-help@r-project.org
> > Asunto: [R] Fast Two-Dimensional Optimization
> >
> >
> > Dear R Helpers,
> >
> > I have attempted "optim" function to solve a two-dimensional
> > optimization problem. It took around 25 second to complete
> > the procedure.
> > However, I want to reduce the computation time: less than 7
> > second. Is there any optimization function in R which is very rapid?
> >
> > Best Regards,
> > Wonsang
> >
> >
> > -----
> > Wonsang You
> > Leibniz Institute for Neurobiology
> > --
> > View this message in context:
> > http://r.789695.n4.nabble.com/R-Fast-Two-Dimensional-Optimizat
> > ion-tp3053782p3053782.html
> > Sent from the R help mailing list archive at Nabble.com.
> >
> >       [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > 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.
> >
>
	[[alternative HTML version deleted]]

Dear Dave,

Thank you for your recommendation. I looked over the website of AD Model
Builder for fast optimization. I want to ask you whether the software can be
used in R programming as a kind of R package? Or is there any possibility
that it can be jointly used with R packages?

Regards,
Wonsang


2010/11/25 dave fournier <davef@otter-rsch.com>
> Hi,
>
> You should be able to get the performance you want with AD Model Builder
> which is free software available at
>
>      http://admb-project.org
>
> I can advise you if you are interested.
>
>      Cheers,
>
>        Dave
>
	[[alternative HTML version deleted]]