Arrays: Multidimensional manipulating

[John W. Eaton]

On 12-Apr-2011, dirac wrote:

| Hi there,
| 
| I am having problems with a multidimensional array at the moment and have
| looked online to see similar problems...
| 
| I basically have an equation
| 
| p = exp(-1*(a*(|k|^b)))
| 
| which I write:
| 
| p = exp(-1*abs(k).^b).
| 
| This is simple enough so far; if I want to vary a, k, and b, I could use for
| loops or I could do it using a multidimensional array (what I want to do as
| I assume it'll be faster). I started by defining vectors for a, b, and k:
| 
| k = [-20:0.05:20];
| b = [0:0.01:2];
| a = [0:0.001:0.01];
| 
| 
| and then tried to find p for all of the possible values of a, b and k and
| store in an array of three dimansion; I tried this:
| 
| p = zeros(length(a),length(k),length(b));   %allocating memory
| 
| p = exp(-1.*a.*(abs(k)).^b)
| 
| Which doesn't seem to work. 
| Is this something that you can do with Octave or will I have to use for
| loops?

  a = 0:0.001:0.01;
  k = -20:0.05:20;
  b = 0:0.01:2;

  na = numel (a);
  nk = numel (k);
  nb = numel (b);

  A = repmat (reshape (a, na, 1, 1), [1, nk, nb]);
  K = repmat (reshape (k, 1, nk, 1), [na, 1, nb]);
  B = repmat (reshape (b, 1, 1, nb), [na, nk, 1]);

  P = exp (-A.*(abs (K)).^B);

  ploop = zeros (na, nk, nb);
  for ii = 1:na
    for jj = 1:nk
      for kk = 1:nb
        ploop(ii,jj,kk) = exp (-a(ii).*(abs (k(jj))).^b(kk));
      endfor
    endfor
  endfor

  absdiff = ploop - P;
  max (absdiff(:))

jwe