一种改进的Hopfield神经网络对TSP问题的求解方法

论文对Hopfield神经网络的能量函数进行重构,使得新能量函数具有参数少、表达式简洁、计算效率高等特点;并引入遗传算法中的变异算子,使得改进后的Hopfield神经网络的具有自适应调整的功能.同时,针对有效解易陷入局部极小值等问题,运用数据转换技术、贪心算法等对有效解进行优化.最后对不同规模的TSP问题仿真,结果表明这些改进方法和技巧是可行的.

A Solving Method to TSP Based on Improved Hopfield Neural Networks

A new type neural networks energy function of HNN （Hopfield neural networks） is proposed in this paper. The energy function is thus given in a simpler formula, fewer parameters and higher computing efficiency. Meanwhile, the mutation operator of genetie algorithm is applied in this HNN, which make the HNN can self-adjust under some conditions. Moreover, greedy algorithm and data transformation technique are introduced in this kind HNN, whieh can make HNN escapes from the local minimum points to some extent. Finally, a series of TSP problems are simulated with different number of cities, and the results of these simulations can explain well that these means is effective to HNN.