A neural network for shortest path computation

This paper presents a new neural network to solve the shortest path problem for internetwork routing. The proposed solution extends the traditional si ngle-layer recurrent Hopfield architecture introducing a two-layer architec ture that automatically guarantees an entire set of constraints held by any valid solution to the shortest path problem. This new method addresses som e of the limitations of previous solutions, in particular the lack of relia bility in what concerns successful and valid convergence. Experimental resu lts show that an improvement in successful convergence can be achieved in c ertain classes of graphs. Additionally, computation performance is also imp roved at the expense of slightly worse results.