Re: Recursive find of set of maximum numbers

[Jordi Gutiérrez Hermoso]

On 20 October 2011 03:11, thombark <address@hidden> wrote:
> I have a vector of integers sorted in descending order. I need to extract the
> indexes of the maximum sum of N values, then the next maximum sum etc.
>
> For example, there are 10 integers a(1) to a(10), where a(1) is the largest
> number and a(10) is the smallest. Let's say N=3. The first max sum would
> give me the obvious vector of (1,2,3) then the next max sum would give me
> (1,2,4).  But the next max sum vector could be (1,2,5) or (2,3,4) depending
> on the values in a(). If a is (10000,10,9,8,7...) then (1,2,5) is the
> resulting vector but if a is (100,90,80,70,10...) then (2,3,4) is the
> result. The very last vector to be returned would be (8,9,10).
>
> Is there a built-in function that would cut down the processing time of this
> algorithm?

Well, you realise you're asking for exponentially many values, but if
this really what you want...

   n = 10; k = 3
   v = sort(rand(1,n), "descend");
   idx = nchoosek (1:n, k);
   [~, idx_s] = sort (sum(v(nchoosek (1:n, k)), 2), "descend");
   idx = idx(idx_s,:);

HTH,
- Jordi G