Getting rid of nested loops, follow up

As a follow up to this question for which I got a really great answer:

I have several K x N matrices and I need to calculate new matrices zed and w according to a function which depends on the "rank" in the column if H is sorted descending, i.e. for

     H = [1,2;5,6;3,4];

Column ranks are 2,3,1 for both columns.

I had previously asked about an operation that involved addition of elements and was able to adapt the answer to several of the functions I was working with. In some of the functions, I have division of two elements depending on different ranks, which are then used in w and zed in a product over elements type calculation.

Example code:

%all matrices are KxN 
w = ones(K,N);
zed = ones(K,N);
for n=1:N        
  [gain, idx] = sort(H(:,n),'descend');                     
  for i=1:K            
    for j=1:K
      if j ~= i
        if  j < i                         
          l = L(idx(j),n)/D(idx(i),n);                        
          w(idx(i),n)=w(idx(i),n)*(B(idx(j),n)^2*H(idx(j),n)^2/l)^(-l);
          p = P(idx(j),n)/Z(idx(i),n);
          zed(idx(i),n)=zed(idx(i),n)*(2*t*B(idx(j),n)*H(idx(j),n)/p)^(-p);
        else  % j > i                 
          r = R(idx(j),n)/Z(idx(i),n);
          zed(idx(i),n) = zed(idx(i),n)*(B(idx(j),n)*H(idx(j),n)/r)^(-r);
        end
      end
    end
  end
end

Initially, I thought I could build larger matrices whose columns/rows represented elements of the product, but I can't figure out how to calculate the p,l, and r values as a matrix, and I'm not sure this approach would actually improve performance.

Any solution, help, suggestions, directions, etc. is all really appreciated.