Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.kei.com!eff!news.umbc.edu!haven.umd.edu!darwin.sura.net!news.Vanderbilt.Edu!NewsWatcher!user
From: goldenjb@ctrvax.vanderbilt.edu (jim golden)
Subject: Re: Routing problem
Message-ID: <goldenjb-271094121609@129.59.170.62>
Followup-To: comp.ai.neural-nets,comp.ai.fuzzy
Sender: news@news.vanderbilt.edu
Nntp-Posting-Host: 129.59.170.62
Organization: vanderbilt
References: <NEWTNews.1451.782929071.cjames@onramp.net>
Date: Thu, 27 Oct 1994 16:55:41 GMT
Lines: 19

In article <NEWTNews.1451.782929071.cjames@onramp.net>, Corby James
<cjames@onramp.net> wrote:

> 
>
> 	At this point, since I am new to this area, I am still struggling with 
> a decision about which architecture and learning rules best fit this problem.  
> Hopfield used the travelling salesman as a metric for his architecture, 
> however, after reading his paper, I have doubts as to whether this approach can 
> be used.  The size of database (1-2 Gb) seems to rule Hopfield nets out. 
> 


I'd try a GA solution before a NN solution.  You'll want a larger sampling
space and you'ld have to test a whole bunch of NN before you could
intuitively feel you had an optimal solution.

Jim Golden
goldenjb@ctrvax.vanderbilt.edu
