R help - May 2011 - Use apply on a function with multiple argument including a matrix ?

Hello,

I wrote a function that takes a matrix M and several numeric parameters as
arguments. It returns a square matrix (same number of rows than M). I need
to use this function n times (n>1000) with random parameters. 

I would like to increase the speed of this computation in order to increase
the size of M  I can use as an input. To do so, I would like to use a
function of the apply family but I am not sure whether it is possible in my
case. Have you got any idea ?

I initially thought I could create a list containing the vectors of the 3
random parameters and run lapply. However, in this case I don't know how to
put M as another argument for the function.

Thank you very much for your help,
Cheers,

Sunelav.

To make things clearer; this is the loop I would like to eliminate

# the function

f=function(M,p1,p2,p3) # A matrix and 3 numeric parameters
{....
return(SM) } # a square matrix


# the Loop

SM=f(M,p1,p2,p3)

for (i in 1:10000)
{ p1=runif(..); p2=runif(..); p3=runif(..)
   SM=SM+f(M,p1,p2,p3) }















--
View this message in context:
http://r.789695.n4.nabble.com/Use-apply-on-a-function-with-multiple-argument-including-a-matrix-tp3556193p3556193.html
Sent from the R help mailing list archive at Nabble.com.

On May 27, 2011, at 3:34 PM, sunelav wrote:
> Hello,
>
> I wrote a function that takes a matrix M and several numeric  
> parameters as
> arguments. It returns a square matrix (same number of rows than M).  
> I need
> to use this function n times (n>1000) with random parameters.
Just one matrix M and many parameters?
>
> I would like to increase the speed
Loops are not necessarily slow in R. Pre-allocation of target object  
size and proper use of indexing and vectorisation may add considerably  
to efficient execution. In a task of this size (which I would call  
"small"), it would appear to be inefficiency in the function that is  
the root of your problem.
> of this computation in order to increase
> the size of M  I can use as an input. To do so, I would like to use a
> function of the apply family but I am not sure whether it is  
> possible in my
> case. Have you got any idea ?
?mapply

mapply("f", p1, p2, p3, moreArgs=list(M=M) )

But no promises about improving on speed versus a loop.
>
> I initially thought I could create a list containing the vectors of  
> the 3
> random parameters and run lapply. However, in this case I don't know  
> how to
> put M as another argument for the function.
>
> Thank you very much for your help,
> Cheers,
>
> Sunelav.
>
> To make things clearer; this is the loop I would like to eliminate
>
> # the function
>
> f=function(M,p1,p2,p3) # A matrix and 3 numeric parameters
> {....
> return(SM) } # a square matrix
>
>
> # the Loop
>
> SM=f(M,p1,p2,p3)
>
> for (i in 1:10000)
> { p1=runif(..); p2=runif(..); p3=runif(..)
>   SM=SM+f(M,p1,p2,p3) }
>
>
>
> View this message in context:
http://r.789695.n4.nabble.com/Use-apply-on-a-function-with-multiple-argument-including-a-matrix-tp3556193p3556193.html

David Winsemius, MD
West Hartford, CT

Dear David,

Thank you very much for your answer. I am going to try what you propose and
compare the two methods. 

I have already done what I could to reduce the computation time of the
function by using apply functions and by doing most of the calculation on
matrices at once to avoid the use of for loops, which I guess is what you
call "vectorisation".  However I have no idea about what you mean by
the
"proper use of indexing". Could you please give me more details about
that ?

Thanks again for your response,

Sunelav



--
View this message in context:
http://r.789695.n4.nabble.com/Use-apply-on-a-function-with-multiple-argument-including-a-matrix-tp3556193p3556475.html
Sent from the R help mailing list archive at Nabble.com.