Generalization and Efficient Implementation of CC4 Neural Network

Abstract

The prescriptive learning and generalization capabilities of the instantaneously trained neural networks make them suitable for various applications. Although the CC4 network, which is the most popular of the instantaneously trained neural networks, does well in comparison with the back propagation network, it suffers from several implementation issues when applied to large data sets. To address these issues, we present a generalized CC4 network. We also describe single neuron implementations of the new network which improve the flexibility and performance of the network. We present results of applications of this network to image processing and time-series prediction.