| This thesis presents a Kohonen Self-Organizing neural network based heuristic for the Traveling Salesman Problem (TSP). A brief overview about neural networks and a review of exact and approximate algorithms for the TSP are given. In order to draw conclusions about the effectiveness of the neural network based heuristic in solving the TSP, the heuristic is used to solve many test problems. An experiment is also performed in order to improve the existing heuristic. With the results of the experiment, a modified heuristic is proposed and tested on the same set of test problems, and solutions are compared to the solutions of the existing heuristic, the optimal solutions, and solutions of other solution techniques. The results of the comparison show that the modified heuristic (like the existing heuristic) possesses many advantages for solving the TSP such as easy implementation, fast computation, robust to test problems, and production of good solutions. The results of the comparison also show that the modified heuristic only used 64% of the run time used by the existing heuristic, and still produced better solutions than those of the existing heuristic. |
