A single neuron model for pattern classification

Abstract

A biologically realistic non linear integrate and fire model is proposed in this paper. Its complete solution is derived and used for the construction of aggregation function in Multi layer perceptron model for classification of UCI Machine learning datasets. It is found that a single neuron in the conventional neural network is sufficient for the classification datasets. It has been observed that the proposed neuron model is far superior in terms of classification accuracy when compared with single integrate and fire neuron model. It is observed that biological phenomenon makes artificial neural network efficient for the classification. © 2012 Springer-Verlag.