Dimensional problem in for loop.

Hi all,

Imagine there is function which takes a matrix as input, perform some operations to each element, and gives an output of same dimension matrix. The dimension of the input matrix can be random. For example:

clear; clc;
% the 1d case.
inpt1d = [1 2 3]';
otpt1d = zeros(3, 1);
for i = 1:3
    % the operation is square root.
    otpt1d(i) = sqrt(inpt1d(i));
end

% the 2d case.
inpt2d = [1 2 3; 4 5 6; 7 8 9]';
otpt2d = zeros(3, 3);
for i = 1:3
    for j = 1:3
        otpt2d(i, j) = sqrt(inpt2d(i, j));
    end
end

Here in the example I manually set i and j for 2d case. However if I want the input to suit any dimension, I cannot set the index manually any more. So for this kind of problem, if I'd like to deal with any dimensions with for loop, how should I do it?

Thanks!

[Greg]

MATLAB inherently understands linear-indexing, do some Google searching on those terms and you should learn you only ever need a single loop (with maybe a tiny bit of slick pre-allocation).

Answer 1

[KL]

EDITED

First you should make sure if loop is required at all in the first place, if yes then I would suggest linear indexing and reshape.

https://www.mathworks.com/help/matlab/math/matrix-indexing.html

https://www.mathworks.com/help/matlab/ref/reshape.html

For example

 A = rand(3,3,3);
 B = A(:);

now it's just a column vector. you could use a single loop with numel(). and once you are done, you could reshape it like

 C = reshape(B,size(A))

[Greg]

I think you missed the question. He is looking for an N-D nested loop solution.

[KL]

Ah yes, I totally misread it, thanks Greg.

[Greg]

The "tiny bit of slick pre-allocation" I referred to in my comment to the original question negates the reshape:

A = rand(3,3,3);
B = 0.*A; % or B = NaN.*A;