Summary: Technische Universit¨at Hamburg­Harburg
Technische Informatik I
Upper Bounds for the Capacity of Structured Neural
Network Architectures
Interner Bericht 6/95
Peter Rieper y , Sabine Kr¨oner z; 1 and Reinhard Moratz zz
y
FB Mathematik, Universit¨at Hamburg, Bundesstr. 55, D­20146 Hamburg
z
Technische Informatik I, TU Hamburg­Harburg, D­21071 Hamburg
zz
AG Angewandte Informatik, Universit¨at Bielefeld, Postfach 100131, D­33501 Bielefeld
July 1995
1 Corresponding author, e­mail: kroener@tu­harburg.d400.de

Abstract
The capacity or Vapnik­Chervonenkis dimension of a feedforward neural architecture is the
maximum number of input patterns that can be mapped correctly to fixed arbitrary outputs.
Moreover it is correlated with the generalization ability of the underlying network architecture,
both important measures for the network design in special applications. So far it is known for

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

Collections: Computer Technologies and Information Sciences