Abstract
ABSTRACT This study is aimed to make a comparison between three artificial neural networks, these networks differ from each other in architecture and the method of adaptive the weights. In this research four ANN are used to recognized English number, these ANN are Adaline, Backpropagation, Hopfield, and Kohen ANN. By doing the comparison, we found that, the ability of a network differentiation does not depend on the complexity of the network architecture, the training algorithm or the number of layers, but it depends on the learning rule and increase in the number of the patterns that are used to train the network.
