"Alex" wrote in message <itvihk$hdn$1@newscl01ah.mathworks.com>...
> I have a large martix with varying number of rows and columns. I want to add all the possible cominations together. I.e all the cominations on a lock:
> say the lock has three columns and 3 rows: (matrix is called array)
> 1 2 3
> 4 5 6
> 7 8 9
>
>
> I have found the function shared on matlab allcomb which can do this if the rows of the array are known.
> Each row can be refered to as array (ii,:) in a for loop
>
> The function allcomb is used in the following fashion
>
> allcomb([1 2 3],[4 5 6],[7 8 9])
> Instead I want to put
> allcomb(array(ii,:),array(ii+1,:),array(ii+2,:))
>
> I.e dynamic values going to the function. This works fine,
> but what if i had a huge matrix. I dont want to have to put in every bit into the function, so can I send it all the rows of the matrix in a for loop or something, because otherwise this would have to be changed if the matrix changed size?
>
> P.S I then want to sum all the rows of combinations. Just saying this incase it affects the method I should use to do this.
>
> Thanks everyone
            
Denote by A your matrix in which we are to select every possible combination of one element from each of its rows.
[m,n] = size(A);
AT = A.';
B = zeros(n^m,m);
B(1:n,m) = AT(:,m);
p = n;
for k = m1:1:1
[I,J] = meshgrid(1:n,1:p);
p = p*n;
B(1:p,k:m) = [AT(I,k),B(J,k+1:m)];
end
Then the n^m by m array B is the desired result. Each row has one of the possible combinations of a single element chosen from each row of A.
I doubt if its speed can compare with that of 'allcomb' but at least it will get you there.
Note: Be careful with your selection of sizes. The quantity n^m can quickly become very large.
Roger Stafford
