A generalized feedforward neural network classifier

In this article a new generalized feedforward neural network (GFNN) architecture for pattern classification is proposed. The GFNNs are an expansion of shunting inhibitory artificial neural networks (SIANNs), proposed previously for classification and function approximations. The GFNN architecture uses as its basic computing unit the generalized shunting neuron (GSN), which includes as special cases the perceptron and the shunting inhibitory neuron. Generalized shunting neurons are capable of forming complex, nonlinear decision boundaries. This allows the GFNN architecture to learn complex pattern classification problems using few neurons. In this article, GFNNs are applied to several benchmark classification problems, and their performance compared to the performance of SIANNs and multilayer perceptrons.

[1]  Stephen Grossberg,et al.  Contour Enhancement, Short Term Memory, and Constancies in Reverberating Neural Networks , 1973 .

[2]  S. Grossberg.  Contour Enhancement , Short Term Memory , and Constancies in Reverberating Neural Networks , 1973 .

[3]  R. B. Pinter.  Product term nonlinear lateral inhibition enhances visual selectivity for small objects or edges. , 1983, Journal of theoretical biology.

[4]  Takayuki Ito,et al.  Neocognitron: A neural network model for a mechanism of visual pattern recognition , 1983, IEEE Transactions on Systems, Man, and Cybernetics.

[5]  R. B. Pinter.  Adaptation of receptive field spatial organization via multiplicative lateral inhibition. , 1984, Journal of theoretical biology.

[6]  Stephen Grossberg,et al.  A massively parallel architecture for a self-organizing neural pattern recognition machine , 1988, Comput. Vis. Graph. Image Process..

[7]  R. B. Pinter,et al.  Shunting inhibitory cellular neural networks: derivation and stability analysis , 1993 .

[8]  A. Bouzerdoum.  The elementary movement detection mechanism in insect vision , 1993 .

[9]  Mohammad Bagher Menhaj,et al.  Training feedforward networks with the Marquardt algorithm , 1994, IEEE Trans. Neural Networks.

[10]  Emile Fiesler,et al.  High-order and multilayer perceptron initialization , 1997, IEEE Trans. Neural Networks.

[11]  Catherine Blake,et al.  UCI Repository of machine learning databases , 1998 .

[12]  Abdesselam Bouzerdoum.  A new class of high-order neural networks with nonlinear decision boundaries , 1999, ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378).

[13]  Abdesselam Bouzerdoum.  Classification and function approximation using feed-forward shunting inhibitory artificial neural networks , 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium.

[14]  Abdesselam Bouzerdoum,et al.  Application of shunting inhibitory artificial neural networks to medical diagnosis , 2001, The Seventh Australian and New Zealand Intelligent Information Systems Conference, 2001.

[15]  James C. Bezdek,et al.  Fuzzy c-means clustering of incomplete data , 2001, IEEE Trans. Syst. Man Cybern. Part B.

[16]  Abdesselam Bouzerdoum,et al.  Expanding the structure of shunting inhibitory artificial neural network classifiers , 2002, Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290).

[17]  R. B. Pinter.  Adaptation of spatial modulation transfer functions via nonlinear lateral inhibition , 1985, Biological Cybernetics.

[18]  S. Grossberg,et al.  Adaptive pattern classification and universal recoding: I. Parallel development and coding of neural feature detectors , 1976, Biological Cybernetics.

[19]  G. G. Furman,et al.  Comparison of models for subtractive and shunting lateral-inhibition in receptor-neuron fields , 1965, Kybernetik.