Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!newsfeed.internetmci.com!news.netrail.net!barney.gvi.net!redstone.interpath.net!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Advantages of NN over Stat. Regression?
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <DJwn01.Gnu@unx.sas.com>
Date: Wed, 20 Dec 1995 21:43:13 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References: <4aqkac$nud@pipe5.nyc.pipeline.com> <mbonnice-1512951629020001@asbe24.phx1.aro.allied.com>
Organization: SAS Institute Inc.
Lines: 19


In article <mbonnice-1512951629020001@asbe24.phx1.aro.allied.com>, mbonnice@pelab.allied.com (Michael A. Bonnice) writes:
|> ...
|> I like neural nets better because they don't complain about singularities
|> and zero determinants.  Neural nets are more robust.

You will have the same problems with singularities with neural nets
as with other varieties of regression. For example, if your inputs
are singular, you cannot assess the relative importance of the
variables involved in the singularity, and you cannot generalize
outside the subspace spanned by the input data. Most statistical
software will tell you when these problems exist. Ignoring such
problems will not make them go away.

-- 

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.
