一种基于量子粒子群的过程神经元网络学习算法

针对过程神经元网络模型学习参数较多,正交基展开后的BP算法计算复杂、不易收敛等问题,提出了一种基于双链结构的量子粒子群学习算法.该算法用量子比特构成染色体,对于给定过程神经元网络模型,按权值参数的个数确定量子染色体的基因数并完成种群编码,通过量子旋转门和量子非门完成个体的更新与变异.算法中每条染色体携带两条基因链,提高了获得最优解的概率,扩展了对解空间的遍历,从而加速过程神经元网络的优化进程.将经过量子粒子群算法训练的过程神经元网络应用于Mackey-Glass混沌时间序列和太阳黑子预测,仿真结果表明该学习算法不仅收敛速度快,而且寻优能力强.

A Learning Algorithm of Process Neural Network Based on Quantum Particle Swarm

Aiming at the problems of high computational complexity and convergence diffculty of the BP(backpropagation) algorithm based on orthogonal basis expansion because there are many parameters in the training of process neural network,a learning method of quantum particle swarm algorithm with double-chain structure is presented.The algorithm uses quantum bits to construct chromosomes.For the given model of process neural network,the number of genes on a chromosome is determined by the number of weight parameters and the population coding is completed.Individuals in the population are updated by quantum rotation gate and mutated by quantum non-gate.In the algorithm,each chromosome carries double chains of genes,which improves the possibility of optimums,expands the traverse of solution space and accelerates the optimization process of the process neural network.The process neural network trained by quantum particle swarm algorithm has been applied to Mackey-Glass time series and the sunspot prediction.The simulation results show that the algorithm not only has the fast convergence but also has good optimization ability.