Dynamical random neural network approach to the traveling salesman problem

Neural networks have been suggested as tools for the solution of hard combinatorial optimization problems. The traveling salesman problem (TSP) is commonly considered as a benchmark for connectionist methods. Here we use the random neural network (RN) model, and apply the dynamical random neural network (DRNN) approach to solve approximately TSP. The advantage of the RN model is that a relatively fast, and purely analytical and numerical approach can be used. Furthermore the RN model equations can be directly solved in full parallelism. We show that DRNN yields solutions of TSP close to the optimal in a majority of the instances tested.<
>