Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!news.ultranet.com!news.sprintlink.net!howland.reston.ans.net!swrinde!elroy.jpl.nasa.gov!sdd.hp.com!hp-pcd!hpcvsnz!briand
From: briand@cv.hp.com (Brian Dixon)
Subject: Re: What Software Package Should I use ??
Sender: news@hpcvsnz.cv.hp.com (News )
Message-ID: <D9t6s8.7Lv@hpcvsnz.cv.hp.com>
Date: Wed, 7 Jun 1995 15:24:08 GMT
References: <3pvg41$rop@Mars.mcs.com>
Nntp-Posting-Host: hpcvsgen.cv.hp.com
Organization: Hewlett-Packard
X-Newsreader: TIN [version 1.1 PL9.4]
Lines: 28

Donald Tveter (drt@MCS.COM) wrote:
: In article <3pt56t$5es@news.primenet.com> you write:
: >What software/hardware package would anyone recommend for a large
: >dataset (300,000+ records)? Brainmaker just isn't doing the job.

: Well I've got a free and a better not free version available for ftp
: and there are other free programs listed in the FAQ you may want to try,
: however I would say that 300,000 records is a bit much for a PC.  One
: of the tricks of training is to give the network a small but
: representative subset (try clustering) and train on that.  Then hit
: the program with the full data set.  Done this way networks have been
: known to learn data that can't be learned with the full data set only
: approach.  What I want to say is that there's a kind of magic in this
: approach by itself, its not just a way to speed up training by cutting
: down on data.  Preparing (pre-processing) data in an appropriate way may
: also help.

Suggested readings:

Learning from Hints in Neural Networks, Yaser S. Abu-Mostafa, Journal of
Complexity #6, 192-198, (1990)

Learning from Hints, Yaser S. Abu-Mostafa, Journal of Complexity #10, 165-178,
(1994)

--
Brian Dixon, Machine Vision Engineer, Hewlett Packard (Corvallis, Oregon)
503-715-3143 (wk), briand@cv.hp.com (email). "Opinions & attitudes are mine!"
