Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!godot.cc.duq.edu!newsgate.duke.edu!zombie.ncsc.mil!news.mathworks.com!newsfeed.internetmci.com!realtime.net!news.mindspring.com!snooze.ser.bbnplanet.com!es.dupont.com!esds01.es.dupont.com!slivova.es.dupont.com!owens
From: owens@slivova.es.dupont.com (Aaron J. Owens)
Subject: Re: ideas about small datasets??
Message-ID: <1996May21.142503.21555@es.dupont.com>
Sender: news@es.dupont.com (USENET News System)
Nntp-Posting-Host: slivova.es.dupont.com
Organization: DuPont Experimental Station
X-Newsreader: TIN [version 1.2 PL0]
References: <DrpAo9.BIr@solair1.inter.NL.net>
Date: Tue, 21 May 1996 14:25:03 GMT
Lines: 35

Ylva Poelman (poelman@evermind.nl) wrote:
: I'm trying to build a neural network for a problem I'm dealing with
: but the trouble is: I have a small dataset to train the network with.
: The networks I've tried didn't learn very much. Does anyone have a
: good suggestions how to tackle this problem, e.g. which network to
: use, which data pretreatment to use etcetera?? Or is it just hopeless
: if you have a smaal dataset??

: Any suggestion is welcome. Thanks.

: Ylva Poelman


Often a "small" data set includes many inputs. Statistical dimensionality
reduction techniques (Principal Component Analysis) can then be applied
to reduce the input dimension and improve the predictive accuracy of the
subsequent neural model. Dimensionality reduction can also be achieved
with an auto-associative neural network model, N -> M -> N, where the
input data X (of dimension N) are used both as the input and target of
the neural network, and there is a "bottleneck" hidden layer of size
M << N.

-- Aaron --

Aaron J. Owens, Research Fellow    Telephone Numbers:
Modeling and Simulation		  	Office	  (302) 695-7341 
Engineering Research Laboratory	  	FAX	    "   695-9658
DuPont Company, E320/201	        Home      (302) 733-7836
Wilmington, DE 19880-0320	   Internet: owens@prism.es.dupont.com

-----------------------------------------------------------------------------
Opinions expressed in this electronic message should *NOT* be taken to repre-
sent the official view(s) of the DuPont Company.  ANY OPINIONS EXPRESSED ARE
THE PERSONAL VIEWS OF THE AUTHOR ONLY. 
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
