Re: Smart sorting algorithm ?

[achrist@easystreet.com]

If you have to invent something to improve the process, where
comparisons are done by humans according to rules unknown, then
there is perhaps some chance to improve the overall speed.  Can 
you give the humans lists of 5-20 items at a time to sort, and
use the information from these sorted sub-lists in some near 
optimal way to produce an overall sort order?   If there is any
machine algorithm that can get you into the neighborhood and put an
approximately sorted list of a few items on  a screen for the person 
to re-arrange into a correctly sorted list,  you should be able to
extract information from the human much faster than you would be
getting it by presenting simple pairwise comparisons.  

The optimal algorithm when comparisons are m-way like that IDK, but 
coming up with one should give lots of opportunity for you to be
creative.


Al 

Wes Groleau wrote:
> 
> > Once again, this is a dead-end idea. It cannot possibly
> 
> Well, I'm not convinced--although I do respect
> your higher-than-mine education and experience
> in this area.  If you provide a "proof" or
> a link to a proof, I'll try to understand it.
> Intuitively, I get foggy images of why it might
> be true, but the wishful thinking is quite
> powerful in this case.
> 
> > be any help if you are using a good sorting algorithm
> > that minimizes comparisons in the first place, and if
> 
> I am convinced of the value of such,
> but since that is not enough in this case......
> 
> --
> Wes Groleau
> http://freepages.rootsweb.com/~wgroleau