Help needed vectorizing layer-wise 3d logical indexing problem.

Hi folks,

I currently have a 3D logical array and a 2D matrix and I would like to logically index the 2D matrix using each layer of the logical array. I was wondering whether there was a faster, possibly more vectorized way that avoids a for loop.

eg.

A is p x q

B is p x q x r

C is cell(1,r)

for i = 1:r
  C{i} = A(B(:,:,i));
end

Is there a one liner that can do this. My motivation is that I may want to parallelize this in the future.

Answer 1

[Mohammad Abouali]

% Creating Sample A and B matrix
A=rand(3,4);
B= (rand(3,4,5))>0.5;

% one liner equivalent to your code.
C=mat2cell(A(mod(find(B)-1,numel(A))+1), ...
           sum(reshape(B,[],size(B,3))))

You have to check though to see if it helps. Sometimes, looping is OK.