Solution One: Vectorized
Your example uses rand, which lets you generate the whole 3D array directly by specifying the outputs size in a vector, no loops or concatenation is required:
out = rand([2,5,3]);
This will be much faster than using any loop and concatenating arrays, and is a good introduction to vectorized code! However if you really want to concatenate arrays, then read on...
Solution Two: with for-Loop
For a start there is no reason why you need to use the variable counter when you already have the for-loop variable. Why bother? And preallocating the array will make this faster. But the basic issue is that you are trying to concatenate and save the matrix in each loop iteration. Why not just create the whole array, and then save it?:
>> N = 3;
>> X = nan([2,6,N]);
>> for k = 1:N
X(:,:,k) = rand(2,6);
end
>> save('test.mat','X')
Bonus Information on Concatenation
>> A = [1,2;3,4];
>> B = [10,20;30,40];
>> C = [0,NaN;Inf,-Inf];
>> out = cat(3,A,B,C)
out(:,:,1) =
1 2
3 4
out(:,:,2) =
10 20
30 40
out(:,:,3) =
0 NaN
Inf -Inf
Note that this is even easier if you place the variables into a cell array:
>> Z = {A,B,C};
>> out = cat(3,Z{:})
