A comparative study of neural network structures for practical application in a pattern recognition environment

The recognition performance of three different types of neural network involving differing structures and different learning algorithms is compared. The networks are the probabilistic logic node, a neuron configuration using a back error propagation algorithm, and the ART1 neural model. The potential of different neural network types in a common practical recognition task is demonstrated and it is shown how architectures and operational parameters might be adjusted in seeking to improve response. The data set available for experimentation is a collection of digitised unconstrained machine-printed alphanumeric characters extracted from postcodes on envelopes in the mail