Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!purdue!ames!enews.sgi.com!news.mathworks.com!news.sprintlink.net!news-peer.sprintlink.net!interpath!news.interpath.net!news.interpath.net!sas!newshost.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: What levels should I set the threshold of my sigmoid?
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <E4uFw8.3tv@unx.sas.com>
Date: Thu, 30 Jan 1997 23:00:08 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <5ck625$q0l@camel0.mindspring.com>
Organization: SAS Institute Inc.
Lines: 22


In article <5ck625$q0l@camel0.mindspring.com>, congee@mindspring.com (Deric Cheng) writes:
|> I guess if you are reading this post you know what 2 variables I'm
|> talking about.  The one that moves the function in the x direction and
|> the function that squishes the sigmoid or make it into one verticle
|> line. =) I used 100 and it looks like a line on my power mac.  Still
|> can you tell me what the levels are set to?

Typically, you don't set them, you train them. For the usual logistic
sigmoid, a fixed nonzero bias theoretically allows universal
approximation, but I don't know of any practical reason to use a fixed
bias.  See "Why use a bias input?" in the Neural Network FAQ, part 2 of
7: Learning, at ftp://ftp.sas.com/pub/neural/FAQ2.html for more
explanation and a reference.

-- 

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.
 *** Do not send me unsolicited commercial or political email! ***

