基于Hopfield-Tank模型的快速神经网络方法

A Fast Neural Network Method Based on Hopfield-Tank Model

分析了Hopfield－Tank模型在收敛性、稳健性、优化率以及计算速度方面存在的问题，根据外部惩罚函数法的基本思想提出了一种新的基于Hopfield－Tank模型的快速神经网络方法。对TSP的能量函数进行了改进，并对我国31个城市的TSP进行了软件模拟，得出了15640km的最短路径，在收敛性、稳健性、优化率以及计算速度方面的结果都十分满意。

The convergence, robustness, optimum and computing speed of Hopfield-Tank are analyzed. And then, according to extemal penalty function, a new fast neural network algorithm based on Hopfield-Tank model is proposed. The TSP's energy function is also improved. According to the numerical experiment for the TSP of 31 cities of our country, the shortest route (15640km is obtained. In the aspects of convergence, robustness, optimum and computing speed, the algorithm is satisfactory.