Analysis of neural network edge pattern detectors in terms of domain functions

Authors:
Maher I. Rajab;Khalid A. Al-Hindi
Affiliations:
Department of Computer Engineering, University of Umm Al-Qura, Mecca, Kingdom of Saudi Arabia;Department of Computer Engineering, University of Umm Al-Qura, Mecca, Kingdom of Saudi Arabia
Venue:
WSEAS Transactions on Information Science and Applications
Year:
2008

Citing 1
Cited 2

Neural Networks: A Comprehensive Foundation

Neural Networks: A Comprehensive Foundation

Electrical energy consumption forecasting in oil refining industry using support vector machines and particle swarm optimization

WSEAS Transactions on Information Science and Applications
The analysis of industrial product external surfaces: a pattern recognition approach to the situation of imprecise limits

WSEAS TRANSACTIONS on SYSTEMS

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper investigates the analysis of feed-forward BP neural network that has been trained to detect noisy edge patterns, so as to achieve close insight into their internal functionality. The analysis of neural network edge detector's hidden units, as templates, were analysed into three gradient components: low pass or averaging, gradient, and second-order gradients. The weights between NNets hidden units and their output units represent the importance of the hidden unit's edge detection outcome. To this purpose, the elements of the NNets, that have been trained to detect prototype noisy edge patterns with various angle of operation, were analysed in terms of domain functions. The results show that the NNets analysis using the domain functions method could confirms the results of NNets recognition accuracies. Although the work presented only gives some analysis results for the units in the NNets hidden units, it should be clear that a characterization of the neural network as a whole could also be derived from these results.