paper available: Noise Robustness in Multilayer Neural Networks

[Michael Biehl]

FTP-host:       ftp.physik.uni-wuerzburg.de 
FTP-filename:   /pub/preprint/1996/WUE-ITP-96-022.ps.gz

The following manuscript is now available via anonymous ftp:
(See below for the retrieval procedure)

------------------------------------------------------------------
    "Noise Robustness in Multilayer Neural Networks"

     M. Copelli, R. Eichhorn, O. Kinouchi, M. Biehl, 
     R. Simonetti, P. Riegler, and N. Caticha

    Ref. WUE-ITP-96-022

                                Abstract
The training of multilayered neural networks in the presence of different types 
of noise is studied. We consider the learning of realizable rules in nonover-
lapping architectures.  Achieving optimal generalization depends on knowledge
of the noise level, however its misestimation may lead to partial or complete loss
of the generalization ability.  We demonstrate this effect in the framework of
online learning and present the results in terms of noise robustness phase
diagrams. While for additive (weight) noise the robustness properties depend on 
the architecture and size of the networks, this is not so for multiplicative 
(output) noise. In this case we find a universal behaviour independent of the 
machine size for both the tree parity and committee machines.     

---------------------------------------------------------------------

Retrieval procedure:

     unix> ftp  ftp.physik.uni-wuerzburg.de
     Name: anonymous  Password: {your e-mail address}
     ftp>  cd pub/preprint/1996
     ftp>  get WUE-ITP-96-022.ps.gz           (*)   
     ftp>  quit
     unix> gunzip WUE-ITP-96-022.ps.gz
e.g. unix> lp WUE-ITP-96-022.ps                     [8 pages]             

(*) can be replaced by "get WUE-ITP-96-022.ps". The file will then
    be uncompressed before transmission (slow!). 
_____________________________________________________________________ 

-- 

Michael Biehl     
Institut fuer Theoretische Physik              
Julius-Maximilians-Universitaet Wuerzburg
Am Hubland
D-97074 Wuerzburg 

email:     biehl at physik.uni-wuerzburg.de
homepage:  http://www.physik.uni-wuerzburg.de/~biehl 
Tel.:      (+49) (0)931 888 5865   
             "      "    "  5131
Fax :      (+49) (0)931 888 5141