Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!julius.cs.uiuc.edu!apple!agate!shelby!helens!news
From: news@helens.Stanford.EDU (news)
Newsgroups: comp.ai.neural-nets
Subject: Re: Back-Propagation Weight Initialization
Message-ID: <1325@helens.Stanford.EDU>
Date: 7 Dec 90 21:09:19 GMT
Organization: Stanford University
Lines: 11

I have a paper titled "Improving the learning speed of two-layer
networks by choosing the initial values of the adaptive weights" in
the proceedings of the IJCNN, June 1990, San Diego, page III-21.  The
co-author is Dr. B. Widrow.  The paper is applicable to networks with one
hidden layer (such networks have been proven to be universal approximator,
giving a sufficient number of hidden units.)  It describes a technique that 
divides the input space into small regions and assign each unit in the
hidden layer to a region.  The algorithm itself is fairly simple.
(I get a factor of 4 or 5 improvement in learning time over random
initial weights.)
Derrick Nguyen