用随机神经网络优化求解C-TSP

基于动态随机神经网络(DRNN:Dynamical Random Neural Network)求解典型旅行商优化问题TSP(Traveling Salesman Problem),通过简化方程参数的改进算法,针对解决大规模TSP的求解效果在时间以及路径寻优上所存在的问题,提出一种新的分区方案来解决中国31城市的旅行商问题.所获得的最优路径结果与目前公开文献中已有的其他神经网络所解的结果相比较,显示出采用随机神经网络解决多于10个变量TSP问题的优越性.实验结果表明,采用该方法解决31个城市TSP的优化,所得出的最短距离(15 112.7km)比已有5种算法的结果都要少.

Solving C-TSP with random neural network

Based on the algorithm of the typical optimal problems-TSP(Traveling Salesman Problem) with DRNN(Dynamical Random Neural Network), simplifying the parameter in feedback equation, aiming at the problem of time-cost and path-search when solving the large scale TSP, a new project of dividing large scale cities into several sub zones is used to solve the China-TSP problem. The final result is compared with other neural networks in references now available. Random neural network is verified to have the advantage to solve TSP with more than 10 variables. The experiment shows that the final path distant of 31-city-TSP (15 112.7 km) is shorter than the ones of 5 known methods.