Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!newsfeed.internetmci.com!EU.net!peer-news.britain.eu.net!newsfeed.ed.ac.uk!dcs.ed.ac.uk!jura.dcs.ed.ac.uk!iic
From: iic@dcs.ed.ac.uk (Ian Clarke)
Subject: Re: nn and river forecasting
X-Nntp-Posting-Host: jura.dcs.ed.ac.uk
Message-ID: <DKyr93.HFC.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk>
To: "James H. Paul" <jhp>
Sender: iic@jura.dcs.ed.ac.uk (Ian Clarke)
Organization: Department of Computer Science, University of Edinburgh
X-Newsreader: xrn 8.00
References:  <4cuuc9$864@hovis.rdc.noaa.gov>
Date: Wed, 10 Jan 1996 11:43:46 GMT
Lines: 28

In article <4cuuc9$864@hovis.rdc.noaa.gov>, "James H. Paul" <jhp> writes:
> i want to try to forecast river levels using a neural net.
> i've read about some neural net software and wonder if
> anyone can suggest a good inexpensive (freeware?) neural
> net application to do this...preferably on a unix system.
> 

If you can program in C, I reccomend the C Neural Network Toolkit, it
was doing the rounds of some newsgroups a while ago, if not, I can't
help you here.

> the other question concerns inputs into this neural net.
> to start off...precipitation and upstream river levels
> woudl be used.  all the examples and documentation i've 
> seen have inputs between zero and one.  will i have to 
> convert these inputs to this range or is there a way to 
> use any real number
> 
You could in theory use any real number, what is critical is the function
which each node uses to determine how to respond to a given input, you
may be able to select this function, but if not, most are designed to 
accept input between 0 and 1, so you should scale your input information
accordingly (it really should not be that much of an inconvenience). 

-- 
|IAN CLARKE        I.Clarke@sms.ed.ac.uk "They couldn't hit an elephant|
|                  iic@dcs.ed.ac.uk      at this dist.."  - Last words |
|I.Clarke@ed.ac.uk ianc@aisb.ed.ac.uk    of General M Howard, 1918     |
