Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!rutgers!usc!samsung!sol.ctr.columbia.edu!emory!utkcs2!mars!sfp
From: sfp@mars.ornl.gov (Phil Spelt)
Newsgroups: comp.ai.neural-nets
Subject: Backprop Weight Initialization
Summary: Recent reference to sensitivity to initial weights
Message-ID: <1990Dec6.161422.5314@cs.utk.edu>
Date: 6 Dec 90 16:14:22 GMT
Expires: 24 December 1990
References: <16266@imag.imag.fr>
Sender: P. F. Spelt
Reply-To: sfp@mars.epm.ornl.gov (Phil Spelt)
Organization: Oak Ridge National Lab
Lines: 9

I remember seeing a reference to an aritcle which dealt with the effects
of random initialization of weights in a backporpagation network.  I believe
the article reported that it *does* make a difference how the weight matrix 
is initialized.

If anyone knows what this reference is (or any others addressing this problem,
i would appreciate hearing from you, either by email or posting to this BB.

Thanks, Phil Spelt