Sensitivity of feedforward neural networks to weight errors

Author

Stevenson, Maryhelen ; Winter, Rodney ; Widrow, Bernard

Author_Institution

Dept. of Electr. Eng., Stanford Univ., CA, USA

Volume

1

Issue

1

fYear

1990

fDate

3/1/1990 12:00:00 AM

Firstpage

71

Lastpage

80

Abstract

An analysis is made of the sensitivity of feedforward layered networks of Adaline elements (threshold logic units) to weight errors. An approximation is derived which expresses the probability of error for an output neuron of a large network (a network with many neurons per layer) as a function of the percentage change in the weights. As would be expected, the probability of error increases with the number of layers in the network and with the percentage change in the weights. The probability of error is essentially independent of the number of weights per neuron and of the number of neurons per layer, as long as these numbers are large (on the order of 100 or more)

Keywords

error statistics; neural nets; probability; sensitivity analysis; Adaline elements; feedforward neural networks; probability; sensitivity analysis; threshold logic units; weight errors; Feedforward neural networks; Geometry; Helium; Innovation management; Logic; Neural network hardware; Neural networks; Neurons; Sensitivity analysis; Terminology;

fLanguage

English

Journal_Title

Neural Networks, IEEE Transactions on

Publisher

ieee

ISSN

1045-9227

Type

jour

DOI

10.1109/72.80206

Filename

80206