A connectionist approach to SDH bandwidth management

In the present paper the application of neural networks to bandwidth management is discussed. Two neural architectures were examined; the radial basis function network, and a Hopfield network. The radial basis function network has been applied to the task of traffic prediction and the Hopfield network used as a linear programming solver. The two networks can be used to assist with the management of bandwidth in a synchronous transmission network, namely a SONET or SDH network.<
>