Parallelized neural network system for solving Euclidean traveling salesman problem

We investigate a parallelized divide-and-conquer approach based on a self-organizing map (SOM) in order to solve the Euclidean traveling salesman problem (TSP). Our approach consists of dividing cities into municipalities, evolving the most appropriate solution from each municipality so as to find the best overall solution and, finally, joining neighborhood municipalities by using a blend operator to identify the final solution. We evaluate performance of parallelized approach over standard TSP test problems (TSPLIB) to show that our approach gives a better answer in terms of quality and time rather than the sequential evolutionary SOM.