[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Arrays: Multidimensional manipulating

From: William Krekeler
Subject: RE: Arrays: Multidimensional manipulating
Date: Tue, 12 Apr 2011 19:26:52 +0000

-----Original Message-----
From: John W. Eaton [mailto:address@hidden 
Sent: Tuesday, April 12, 2011 11:51 AM
To: William Krekeler
Cc: dirac; address@hidden
Subject: RE: Arrays: Multidimensional manipulating

On 12-Apr-2011, William Krekeler wrote:

| Maybe you could try using meshgrid.
| [A,B,K] = meshgrid(a,b,k);
| p = size(A);
| p = exp(-1.*A.*(abs(K)).^B);

What is the second line for?

Using meshgrid is good if you need exactly three dimensions.  If you
need more, then you will need something like the method I showed with
repmat and reshape.  But the basic idea is the same.  You need to
generate a set of matrices all with the same size that together
contain all the combinations that you are looking for.  Of course if
you need many dimensions, you will quickly run out of memory storing
the temporary matrices that represent all combinations...


I meant to type 'p = zeros( size(A) );' to preallocate the array. Either way it 
is an unnecessary step.

I agree that using meshgrid will result in increased memory usage, but for the 
small matrices used I think (untested) that the meshgrid vector math solution 
may be faster than a sequence of for loops. Vector computations is generally 
faster than for loops, at least in base Octave without compilation. As Octave 
does not benefit from for loop acceleration like Matlab's locally compiled just 
in time for loop implementation.

William Krekeler

reply via email to

[Prev in Thread] Current Thread [Next in Thread]