Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!pasteur!ucbvax!decwrl!labrea!glacier!jbn
From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.ai.neural-nets
Subject: Re: implementation of "traveling salesman algorithm" on connection machine
Keywords: TSP, connection machine
Message-ID: <18079@glacier.STANFORD.EDU>
Date: 3 Feb 89 17:28:46 GMT
References: <1637@cps3xx.UUCP>
Reply-To: jbn@glacier.UUCP (John B. Nagle)
Distribution: usa
Organization: Stanford University
Lines: 9


      There's a lot of hype about this.  The problem isn't that hard if you
are willing to settle for a result that's almost optimal, rather than going
for the true optimum, which is what the "neural" approaches do anyway.
The business about it taking hours on a mainframe to solve a 30-node net
is pure bunk.  It takes less than 5 seconds on a PC/AT, using the 1970-vintage
Bell Labs algorithm.

					John Nagle