基于混合量子遗传算法的过程神经元网络训练 *

针对过程神经元网络现有学习算法复杂度高、对初值敏感的问题，提出了一种基于混合量子遗传算法 的过程神经元网络训练方法。将过程神经元网络的训练转换为等价非线性方程组的优化求解问题，用量子比特 构成染色体，采用实数对染色体进行编码，同时引入拟牛顿算法作局部搜索。该算法可发挥量子遗传算法的群 体搜索能力和全局收敛性，以及拟牛顿法较快的收敛速度，同时有效克服了拟牛顿算法对初值敏感的问题。训 练结果表明，此算法具有较好的稳定性和有效性。

Training method of process neural networks based on hybrid quantum genetic algorithm

Aiming at the problems that current learning algorithms of PNN were highly complex and sensitive to initial values,this paper presented a training method of PNN based on hybrid quantum genetic algorithm.This method transformed the training of PNN into the optimization solution of nonlinear equations,and used quantum bits to construct chromosomes which were coded with real number.Meanwhile,introduced was to implement local searching.The algorithm had the abilities of population search for quantum genetic alg...