An Accurate Measure for Multilayer Perceptron Tolerance to Weight Deviations

Authors:
Jose L. Bernier;J. Ortega;M. M. Rodrì‘guez;I. Rojas;A. Prieto
Affiliations:
Dpto. Arquitectura y Tecnologìa de Computadores, Universidad de Granada, E18071--Granada, Spain, e-mail: jbernier@atc.ugr.es, aprieto@atc.ugr.es;Dpto. Arquitectura y Tecnologìa de Computadores, Universidad de Granada, E18071--Granada, Spain, e-mail: jbernier@atc.ugr.es, aprieto@atc.ugr.es;Dpto. Arquitectura y Tecnologìa de Computadores, Universidad de Granada, E18071--Granada, Spain, e-mail: jbernier@atc.ugr.es, aprieto@atc.ugr.es;Dpto. Arquitectura y Tecnologìa de Computadores, Universidad de Granada, E18071--Granada, Spain, e-mail: jbernier@atc.ugr.es, aprieto@atc.ugr.es;Dpto. Arquitectura y Tecnologìa de Computadores, Universidad de Granada, E18071--Granada, Spain, e-mail: jbernier@atc.ugr.es, aprieto@atc.ugr.es
Venue:
Neural Processing Letters
Year:
1999

Citing 4
Cited 6

Adaptive fuzzy systems and control: design and stability analysis

Adaptive fuzzy systems and control: design and stability analysis
Comparative Fault Tolerance of Parallel Distributed Processing Networks

IEEE Transactions on Computers
A Modified Backpropagation Algorithm to Tolerate Weight Errors

IWANN '97 Proceedings of the International Work-Conference on Artificial and Natural Neural Networks: Biological and Artificial Computation: From Neuroscience to Technology
Complete and partial fault tolerance of feedforward neural nets

IEEE Transactions on Neural Networks

Obtaining Fault Tolerant Multilayer Perceptrons Using an Explicit Regularization

Neural Processing Letters
Assessing the Noise Immunity and Generalization of Radial Basis Function Networks

Neural Processing Letters
A Quantitative Study of Fault Tolerance, Noise Immunity, and Generalization Ability of MLPs

Neural Computation
Analysis on Bidirectional Associative Memories with Multiplicative Weight Noise

Neural Information Processing
On objective function, regularizer, and prediction error of a learning algorithm for dealing with multiplicative weight noise

IEEE Transactions on Neural Networks
On the selection of weight decay parameter for faulty networks

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

The inherent fault tolerance of artificial neuralnetworks (ANNs) is usually assumed, but severalauthors have claimed that ANNs are not always faulttolerant and have demonstrated the need to evaluatetheir robustness by quantitative measures. For thispurpose, various alternatives have been proposed. Inthis paper we show the direct relation between themean square error (MSE) and the statisticalsensitivity to weight deviations, defining a measureof tolerance based on statistical sentitivity that wehave called Mean Square Sensitivity (MSS); this allowsus to predict accurately the degradation of the MSEwhen the weight values change and so constitutes auseful parameter for choosing between differentconfigurations of MLPs. The experimental resultsobtained for different MLPs are shown and demonstratethe validity of our model.