A Neural Routing Strategy which Adapts to Changes in Computer Network Topology

In the paper a neural approach to routing in computer network is proposed. It is based on the use of a Hopfield-type neural model in the switch nodes, which, on the basis of a global vision of the network, deterniines the optimal path to each destination. An essential requisite in routing is the capacity to adapt to variations in the computer network topology. In the paper, the authors demonstrate the capacity of the neural solution to adapt to topological variations as well, thus enhancing its applicability to real cases.