Feedback Neural Networks

Abstract

The artificial neural networks discussed in this chapter have different architecture from that of the feedforward neural networks introduced in the last chapter. That is, there are inherent feedback connections between the neurons of the networks. For the feedforward neural networks, such as the simple or multilayer perceptrons, the feedback-type interactions do occur during their learning, or training, stage. Information about the weight adjustment is fed back to the various layers from the output layer to reduce the overall output error with regard to the known input-output experience. When the training stage ends, the feedback interaction within the network no longer remains.