Application of Hopfield neural network in routing for computer networks

In packet switched computer networks, data are routed along communication links making up the network, the problem arises of how to determine the shortest paths to route the data. Each link has a cost representing the desirability of using that particular link. The shortest path then becomes the path in which the total link cost from a source node to a destination node is minimized. While there are many conventional algorithms available, non of them are without drawbacks. In order to find a better solution, the use of Hopfield neural network as a routing algorithm is explored in this paper. This approach is based on a solution proposed for the Traveling Salesman Problem (TSP). The routing problem is considerably similar to the TSP problem in that it is concerned with finding an optimal route which connects a source node with a destination node. The neural network architecture is implemented and tested, and a comparison between the neural solution and the conventional routing algorithms is also presented. The neural network gives approximately 97.3% of optimum routes, this represents a significant improvement over other neural network approaches already implemented.