Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!satisfied.elf.com!news.mathworks.com!panix!bloom-beacon.mit.edu!news.media.mit.edu!minsky
From: minsky@media.mit.edu (Marvin Minsky)
Subject: Re: TSP (Traveller Saleman Problem)
Message-ID: <1995Feb5.183455.27552@news.media.mit.edu>
Sender: news@news.media.mit.edu (USENET News System)
Cc: minsky
Organization: MIT Media Laboratory
References: <3gr8sc$9t8@news.cict.fr> <3gt1qv$n3e@odin.diku.dk> <3h33gh$r41@news.bu.edu>
Date: Sun, 5 Feb 1995 18:34:55 GMT
Lines: 26

 meissner@space.bu.edu (Karl Meissner) writes:
>Anders Nielsen (joshu@diku.dk) wrote:
>>In <3gr8sc$9t8@news.cict.fr> Cerf Stephane <cerf@int-evry.fr> writes:

>>>I am Looking for the solution at the TSP using GA in C or C++. Does
anybody know the solution or where I can find it. I love GA but I am a
kind of lost in the Goldberg book. Thanks. CIAO.

>> As far as I know, the travelling salesman problem hasn't been
solved. Please correct me if Im wrong.

>Actually TSP has been done for several thousand cities, using GAs.  
>Simulated Annealing has comparable results.  

This thread seems confused about what it means to "solve" TSP.  Many
programs purport to "do well" on TSP problems.  But the original
mathematical problem was to find (A)  *the shortest path* -- not (B) "a pretty
short path".  I think that all these claims are of type (B), while it
has been proved that (A) is NP.  Worse, so far as I know, there is no
"nontrivial several thousand city" example for which the (A) solution
is even known, so that there's no way to tell when you found it.
There's certainly no reason to suspect that GA could find it reliably,
in NP time.  

