Recurrent and feedforward backpropagation for time independent pattern recognition

Based on a unified description of neural algorithms for time-independent pattern recognition, the authors discuss the generalization ability of a three-layer perceptron for recurrent backpropagation depending on the number of learning epochs and the steepness of the neurons' threshold function. Moreover, the suitability of recurrent and feedforward backpropagation learning algorithms for implementation on multiprocessor systems is investigated. In contrast to the common belief that recurrent back-propagation is more computationally intensive than feedforward backpropagation, the present results for optical character recognition indicate that this need not be the case if the steepness of the threshold function is appropriately chosen. This makes recurrent backpropagation a suitable candidate for time-independent pattern recognition.<
>