Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Technische Universitat HamburgHarburg Technische Informatik I
 

Summary: Technische Universit¨at Hamburg­Harburg
Technische Informatik I
A Learning Algorithm for Structured Invariant
Neural Networks
Interner Bericht 5/96
Sabine Kr¨oner
Technische Informatik I
Technische Universit¨at Hamburg­Harburg
D­21071 Hamburg
e­mail: kroener@tu­harburg.d400.de
September 1996

Abstract
Structured multi­layer feedforward neural networks gain more and more importance
in speech and image processing applications. The characteristic of structured neural
networks is that a­priori knowledge about the task to be performed is directly built
into their architecture by the use of nodes with shared weight vectors. Usually these
networks are trained with a modified backpropagation algorithm. However, due to the
constraints on the weight vectors it is difficult to find suitable initialization values, and
often the algorithm fails to converge.

  

Source: Albert-Ludwigs-Universität Freiburg, Institut für Informatik,, Lehrstuhls für Mustererkennung und Bildverarbeitung

 

Collections: Computer Technologies and Information Sciences