A Solution of the Travelling Salesman Problem by a Neural Network with Self-Feedbacks
Hiroaki Yamamoto, Kenji Murakami, Takahumi Oohori, Kazuhisa Watanabe · IEEJ Transactions on Electronics Information and Systems · 1996
A Hopfield network with self-feedbacks for the travelling salesman problem(TSP) is proposed. However, the self-feedback makes the network converge to a solution which is not feasible but has much information about the optimum solution. Therefore, in this paper, the influence of the self-feedback upon the solution is theoretically discussed and the method to improve feasibility by neuron outputs and activities is proposed.Computational results on the TSP with 5_??_30 cities are reported, which show that the proposed method obtains higher performance than the conventional Hopfield method.