ANN with two-dendrite neurons and its weight initialization

  • Download Citations
  • Email
  • Print
  • Rights And Permissions

Access The Full Text

Sign In:Full text access may be available with your subscription

Forgot Username/Password?Athens/Shibboleth Sign In


Chen, Y.;   Bastani, F.;  
Western Geophysical, Houston, TX 

This paper appears in: Neural Networks, 1992. IJCNN., International Joint Conference on
Issue Date: 7-11 Jun 1992
On page(s): 139 - 146 vol.3
Meeting Date: 07 Jun 1992 - 11 Jun 1992
Location: Baltimore, MD , USA
Print ISBN: 0-7803-0559-0
Cited by : 1
INSPEC Accession Number: 4422065
Digital Object Identifier: 10.1109/IJCNN.1992.227179 
Date of Current Version: 06 August 2002

Abstract

The authors propose the multidendrite multiactivation product unit and the vectorial connection model for artificial neural networks. A generalized backpropagation learning rule is also developed for multilayer feedforward networks with a new neuron model and connections. Each hidden neuron is a multiactivation product unit which requires vectorial axon connections and a productive activation function. An optimal weight initialization algorithm is developed for a three-layer network with hidden units of 2D vectorial connections. The weights between the input layer and the hidden layer are derived from the feature selection methods used in pattern recognition. The activation function is the product of a 2D Hermite spline base function. The weights between the hidden layer and the third layer are scaled coefficients of the 2D Hermite spline interpolations. The performances of networks initialized by the new algorithm are compared with those obtained by selecting random initial weights

Available to subscribers and IEEE members.

Available to subscribers and IEEE members.

Available to subscribers and IEEE members.



Indexed by Inspec

© Copyright 2012 IEEE – All Rights Reserved