[R] avoiding loops

[Bill.Venables at csiro.au]

Ingmar Visser asks:
> 
> Thanks for this.
> I was afraid someone was going to say this ...
> Does this mean the only way of getting this to run faster is by  
> moving to C code?

Perhaps, but that's not all that difficult for this kind of operation,
surely?  Even portability across platforms should be pretty
manageable.

> The cases I'm thinking of applying this in have dimensions of A that  
> are much larger than
> the example, eg n by n by T where n has a max of 10 or so but T could

> be hundreds
> or even thousands.

What I would try first, then is a loop which makes maximum use of
vectorisation. In this case it would be two loops, of course:

d <- dim(A)
for(i in 1:d[1])
  for(j in 1:d[2])
    A[i,j,] <- A[i,j,] * B

But the following non-loop solution might be your best option:

A <- aperm(aperm(A, c(3,1,2)) * as.vector(B), c(2,3,1))

If you can get your head around working with an A array where what is
now the third dimension becomes the first, then you can eliminate the
aperm()s and it becomes about as fast as C code (which it is, in
effect).  Permuting large arrays can be heavy on memory usage.

Bill Venables.


> Best, Ingmar
> 
> On Mar 27, 2008, at 12:11 AM, <Bill.Venables at csiro.au>  
> <Bill.Venables at csiro.au> wrote:
> 
> > If you have lots of memory there is an obvious strategy:
> >
> > d12 <- prod(dim(A)[1:2])
> > A <- A * array(rep(B, each = d12), dim = dim(A))
> >
> > I don't really see much wrong with the obvious for() loop, though:
> >
> > for(b in 1:length(B)) A[,,b] <- A[,,b] * B[b]
> >
> >
> > Bill Venables
> > CSIRO Laboratories
> > PO Box 120, Cleveland, 4163
> > AUSTRALIA
> > Office Phone (email preferred): +61 7 3826 7251
> > Fax (if absolutely necessary):  +61 7 3826 7304
> > Mobile:                         +61 4 8819 4402
> > Home Phone:                     +61 7 3286 7700
> > mailto:Bill.Venables at csiro.au
> > http://www.cmis.csiro.au/bill.venables/
> >
> > -----Original Message-----
> > From: r-help-bounces at r-project.org [mailto:r-help-bounces at r- 
> > project.org]
> > On Behalf Of Ingmar Visser
> > Sent: Thursday, 27 March 2008 7:58 AM
> > To: R-help at r-project.org
> > Subject: [R] avoiding loops
> >
> > Hi,
> > I need to compute an array from a matrix and an array:
> >
> > A <- array(1:20,c(2,2,5))
> > B <- matrix(1:10,5)
> >
> > And I would like the result to be an array consisting of the  
> > following:
> >
> > rbind(A[1,,1]*B[1,],
> > A[2,,1]*B[1,])
> >
> > rbind(A[1,,2]*B[2,],
> > A[2,,2]*B[2,])
> >
> > rbind(A[1,,3]*B[2,],
> > A[2,,3]*B[2,])
> >
> > etc.
> >
> > Hence the result should have the same dimension as A, ie a series of
> > 2 by 2 matrices.
> >
> > Short of a for loop over the the last index of A I have struggled
> > with versions of apply but
> > to no avail ...
> >
> > Any insights much appreciated,
> >
> > Best, Ingmar
> >
> > ______________________________________________
> > 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.
> >
> >
>