Call the classification capability of network be further improved by usingquadratic sigmoidal neurons?

In Ref. [4] by using constructive method, Chiang et al., proved that a thre e-layer neural network containing k + 1 single threshold quadratic sigmoida l hidden neurons and one multithreshold sigmoidal output neuron could learn arbitrary dichotomy defined on a training set of 4k patterns. In this pape r the classification capability of the feed forward neural networks contain ing multiple or single threshold quadratic sigmoidal neurons in the hidden and output layer is evaluated. The degree of improvement on the classificat ion capability of network by using quadratic sigmoidal neurons is analyzed. Published by Elsevier Science Ltd.