Storage properties of correlated perceptrons

Author: Engel A.   Malzahn D.   Kanter I.  

Publisher: Taylor & Francis Ltd

ISSN: 1463-6417

Source: Philosophical Magazine B, Vol.77, Iss.5, 1998-05, pp. : 1507-1513

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

Feedforward multilayer neural networks implementing random input-output develop characteristic correlations between the activity of their hidden which are important for the understanding of their storage and performance. It is shown how these correlations can be within the replica-symmetric approximation. Replacing the network by an ensemble of perceptrons displaying the same the relative influence of these correlations on the storage capacity be studied.