home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Monster Media 1993 #2
/
Image.iso
/
text
/
9305nni.zip
/
930529.PPR
< prev
next >
Wrap
Internet Message Format
|
1993-05-29
|
3KB
From ml-connectionists-request@q.cs.cmu.edu Sat May 29 03:18:59 1993
Received: by cse.uta.edu (5.57/Ultrix2.4-C)
id AA24055; Sat, 29 May 93 05:18:42 -0500
Received: from Q.CS.CMU.EDU by q.cs.CMU.EDU id aa21860; 29 May 93 4:44:45 EDT
Received: from DST.BOLTZ.CS.CMU.EDU by Q.CS.CMU.EDU id aa21833;
29 May 93 4:22:23 EDT
Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa11190;
29 May 93 4:21:32 EDT
Received: from RI.CMU.EDU by B.GP.CS.CMU.EDU id aa02268; 29 May 93 4:19:15 EDT
Received: from danpost.uni-c.dk by RI.CMU.EDU id aa13039; 29 May 93 4:18:45 EDT
Received: from nn.meatre.dk by danpost.uni-c.dk (5.65c+/1.34)
id AA06535; Sat, 29 May 1993 10:17:45 +0200
Received: by nn.meatre.dk (4.1/SMI-4.1)
id AA01977; Sat, 29 May 93 10:17:44 +0200
Date: Sat, 29 May 93 10:17:44 +0200
From: Hans Henrik Thodberg <thodberg@nn.meatre.dk>
Message-Id: <9305290817.AA01977@nn.meatre.dk>
To: Connectionists@cs.cmu.edu
Subject: Ace of Bayes (TR announcement)
Status: RO
FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/thodberg.ace-of-bayes.ps.Z
The file thodberg.ace-of-bayes.ps.Z (40 pages) is available from Neuroprose:
ACE OF BAYES: APPLICATION OF NEURAL NETWORKS WITH PRUNING
Hans Henrik Thodberg
Abstract
MacKay's Bayesian framework for backpropagation is a practical and
powerful means of improving the generalisation ability of neural
networks. The framework is reviewed and extended in a pedagogical way.
The notation is simplified using the ordinary weight decay parameter,
and the noise parameter $\beta$ is shown to be nothing more than an
overall scale. A detailed and explicit procedure for adjusting several
weight decay parameters is given.
Pruning is incorporated into the Bayesian framework. Appropriate
symmetry factors on sparse architectures are deduced. Bayesian weight
decay is demonstrated using artificial data generated by a sparsely
connected network. Pruning yields computational advantages: by removing
unimportant weights the posterior weight distribution becomes Gaussian,
and pruning removes zero-modes of the Hessian and redundant hidden
units. In addition, pruning improves generalisation. The Bayesian
evidence is used as a stop criterion for pruning.
Bayesian backprop is applied in the prediction of fat content in minced
meat from near infrared spectra. It outperforms ``early stopping'' as
well as quadratic regression. The evidence of a committee of
differently trained networks is computed and the corresponding improved
generalisation is verified. The error bars on the predictions of the
fat content are computed. There are three contributors: The random
noise, the uncertainty in the weights, and the deviation among the
committee members. Finally the Bayesian framework is compared to
Moody's GPE.
-----
The paper is in the compressed Postscript file
thodberg.ace-of-bayes.ps.Z. Hardcopies are not available.
The Postscript file prints out smoothly on the Postscript
printers from HP, Canon, Apple, Sparc etc., but not on a NEC.
Comments are welcome.
------------------------------------------------------------------
Hans Henrik Thodberg Email: thodberg@nn.meatre.dk
Danish Meat Research Institute Phone: (+45) 42 36 12 00
Maglegaardsvej 2, Postboks 57 Fax: (+45) 42 36 48 36
DK-4000 Roskilde, Denmark
------------------------------------------------------------------