how to get/obtain a specific column of a cell array

Hi,

I have the following code, that transforms a 3D array into a cell array. This code was already suggested on the thead: how to convert a 3D array into a n 2D arrays - (mathworks.com)

A=rand(20,1000,30);
for i = 1:20
    B{i} = squeeze(A(i,:,:));
end

the dimension of B is 20 cells of dimension 1000*30 each

I would like to select/get/obtain the column 1, 4 and 5 of each cell (currently 30). How would I do it?

I thank you in advance,

Best regards,

Answer 1

[Jan]

While this is trivial using the original numerical matrix, it is a mess using cells. This shows, that for your problem the decision to use cell is a bad choice.

At least it works:

A = rand(20,1000,30);
B = cell(1, 20);   % Pre-allocate for speed!
for i = 1:20
    B{i} = squeeze(A(i,:,:));
end
B{3}(:, [1,4,5])

To do this for all elements of the cell array, use a loop. It works with cellfun also, but it is slower and worse to read.

[Hugo]

Hi,

Thank you such much for the swift and accurate reply. The problem is that I would like to use matlab functions that do not support 3D arrays. Is there an easier way to transform a 20*1000*30 in 20 arrays of dimensions 1000*30 ? As you can see, I am using squeeze, but I have problems using the resultant cell array in posterior operations.

[Jan]

Why do you want to split the compact and efficient array into elements of a cell? What is the benefit of B{i} compared to A(i, :, :)? If you stay at the original array, the solution would be: A(:, :, [1,4,5]).

By the way, storing 20 matrices might be more efficiently, if you stack them along the 3rd dimension, e.g. as

A = rand(1000, 30, 20);

Then A(:, :, i) is a 2D matrix automatically without calling squeeze, because Matlab let trailing dimensions of size 1 vanish automagically.