R help - Sep 2007 - Passing parameters to 'optim' fn function

Hi again! I'm using the 'optim' method to fix the parameters of a 
model. I have written the function to be minimised and another function 
which returns the gradient of the error. Now my problem is that, in 
order to compute that gradient, it would be extremely convenient to be 
able to pass some parameters to the gradient function. I don't see how 
to do it given the fixed syntax of 'optim', which does not allow for 
parameters being passed to fn and gr:

 > optim(par, fn, gr = NULL, ...)

    Of course the first idea would be to "pack" the extra parameters
in
the vector 'par', but in that case the extra parameters would be 
optimized too.
    Does anyone have an idea on how to pass parameters to gr in optim? 
Thanks for your time!

Hi

Use the dots argument at the end to pass in named parameters.
This is documented as the last in the argument list under ?optim.


Here's a toy example using x and p:

 > f <- function(x,p){sum(x^2)+p*sum(x)}
 > gr <- function(x,p){2*x+p*rep(1,length(x))}
 > optim(par=c(10,10),fn=f,gr=gr,p=13)
$par
[1] -6.498188 -6.499652

$value
[1] -84.5

$counts
function gradient
       63       NA

$convergence
[1] 0

$message
NULL


HTH

rksh


On 12 Sep 2007, at 09:32, Jos? Luis Aznarte M. wrote:
>     Hi again! I'm using the 'optim' method to fix the
parameters of a
> model. I have written the function to be minimised and another  
> function
> which returns the gradient of the error. Now my problem is that, in
> order to compute that gradient, it would be extremely convenient to be
> able to pass some parameters to the gradient function. I don't see how
> to do it given the fixed syntax of 'optim', which does not allow
for
> parameters being passed to fn and gr:
>
>> optim(par, fn, gr = NULL, ...)
>
>     Of course the first idea would be to "pack" the extra  
> parameters in
> the vector 'par', but in that case the extra parameters would be
> optimized too.
>     Does anyone have an idea on how to pass parameters to gr in optim?
> Thanks for your time!
>
> ______________________________________________
> R-help at 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.
--
Robin Hankin
Uncertainty Analyst
National Oceanography Centre, Southampton
European Way, Southampton SO14 3ZH, UK
  tel  023-8059-7743

you need to use the `...' argument; look at ?optim() for more info.

Best,
Dimitris

----
Dimitris Rizopoulos
Ph.D. Student
Biostatistical Centre
School of Public Health
Catholic University of Leuven

Address: Kapucijnenvoer 35, Leuven, Belgium
Tel: +32/(0)16/336899
Fax: +32/(0)16/337015
Web: http://med.kuleuven.be/biostat/
     http://www.student.kuleuven.be/~m0390867/dimitris.htm


----- Original Message ----- 
From: ""Jos? Luis Aznarte M."" <aznarte at ugr.es>
To: <r-help at stat.math.ethz.ch>
Sent: Wednesday, September 12, 2007 10:32 AM
Subject: [R] Passing parameters to 'optim' fn function

>    Hi again! I'm using the 'optim' method to fix the parameters
of a
> model. I have written the function to be minimised and another 
> function
> which returns the gradient of the error. Now my problem is that, in
> order to compute that gradient, it would be extremely convenient to 
> be
> able to pass some parameters to the gradient function. I don't see 
> how
> to do it given the fixed syntax of 'optim', which does not allow
for
> parameters being passed to fn and gr:
>
> > optim(par, fn, gr = NULL, ...)
>
>    Of course the first idea would be to "pack" the extra
parameters
> in
> the vector 'par', but in that case the extra parameters would be
> optimized too.
>    Does anyone have an idea on how to pass parameters to gr in 
> optim?
> Thanks for your time!
>
> ______________________________________________
> R-help at 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.
> 

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

Jos? Luis Aznarte M. wrote:
>     Hi again! I'm using the 'optim' method to fix the
parameters of a
> model. I have written the function to be minimised and another function 
> which returns the gradient of the error. Now my problem is that, in 
> order to compute that gradient, it would be extremely convenient to be 
> able to pass some parameters to the gradient function. I don't see how 
> to do it given the fixed syntax of 'optim', which does not allow
for
> parameters being passed to fn and gr:
>
>  > optim(par, fn, gr = NULL, ...)
>
>     Of course the first idea would be to "pack" the extra
parameters in
> the vector 'par', but in that case the extra parameters would be 
> optimized too.
>     Does anyone have an idea on how to pass parameters to gr in optim? 
> Thanks for your time!
>   
You can put the parameters in the ... argument (watch out for matches to 
earlier optim args!), or into the environments of fn and gr.  For example,

makefn <- function(x) {
  fn <- function(mu) sum((x-mu)^2)
  gr <- function(mu) sum(2*(mu-x))
  list(fn = fn, gr = gr)
}
x <- rnorm(10)
fg <- makefn(x)
optim(1, fg$fn, fg$gr)

Duncan Murdoch