一种求解TSP问题的新型人工神经网络方法

A New Manual Neural Networks Method for Solving TSP

该文基于输出 -阈值耦合神经网络的自动波现象 ,提出了一种用自动波方法求解TSP问题的方法。该方法具有鲁棒性和可靠性好、大规模并行计算等特点 ,可用于求解对称、非对称赋权图的TSP问题。与目前其它求解TSP问题的方法相比 ,自动波方法执行更为简单 ,不需要太多人为的选择参数等问题 ,且不存在局部极小点的问题 ,求得的解全部是最优解。其所需的计算量 (迭代次数 )主要取决于最短回路的长度 ,而与图的复杂程度、所存在的通路总数关系不大。

Based on the autowaves in the output-threshold coupled neural network(OTNN), this paper presents a method for solving Traveling Salesman Problem(TSP) with OTNNs. The method presented here features in good robustness, high reliability, and large scale of parallel computation. Compared with other methods for solving TSP, it is simple and requires less number of parameters necessary to be selected. Its another advantage is that its solutions are always the optimum. It is shown that OTNN is very effective in solving the TSP for symmetric and asymmetric weighted graph, with the number of iterations mostly lying on the length of the shortest loops, rather than the complexity of the graph and the total number of existed paths in the graph. Finally, experiments on solving TSP are presented.