You can't do this with any old NxK matrix, it must be a square NxN matrix. Otherwise your condition, that the sums of individual rows are 1 and sums of individual cols are also 1, will not be valid. There isn't a simple function (that I know of), but there is a simple solution (provided you are not trying to find for large N). See the example below for N = 4:
N = 4;
A = eye(N);
B = perms(1:N);
C = zeros(N,N,size(B,1));
for ii = 1 : size(B,1)
C(:,:,ii) = A(:,B(ii,:));
end
disp(C)
