Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!gatech!taco.cc.ncsu.edu!news-server.ncren.net!redstone.interpath.net!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: NN and regression
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <DM5sv1.7xq@unx.sas.com>
Date: Fri, 2 Feb 1996 17:35:25 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References: <4ddbe0$q9h@nuscc.nus.sg> <3106FF80.38E5@clark.net> <hrLrJcB.laneo@delphi.com>
Organization: SAS Institute Inc.
Lines: 21


In article <hrLrJcB.laneo@delphi.com>, Topspin <laneo@delphi.com> writes:
|> Big difference.  NNs require none of the implicit assumption of regression
|> and that is one of their many benefits. 

NNs involve precisely the same assumptions as linear or logistic
regression except for one: linearity. There is a discussion of this
issue with numerous references in the FAQ. Point your web browser to
ftp://ftp.sas.com/pub/neural/FAQ.html and look under "Part 2: Learning:
How are NNs related to statistical methods?"

|> Moreover, NNs will allow you to use
|> incomplete or noisy data---try that with regression.

Regarding incomplete or noisy data, I don't know of any way in which NNs
work differently than regression.
-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
