摘要
提出一种基于量子激励粒子群算法优化BP网络参数的新方法。该算法在粒子群优化算法中引入量子论思想,克服了传统粒子群算法易陷入局部极值、优化效果较差的不足,最终得到BP网络的最佳参数值。将该算法应用于3个典型复杂函数,并与传统BP算法、基于传统的粒子群优化BP网络算法的仿真结果进行分析对比。结果表明:该算法训练次数少,模型精度高,性能优于其它两种BP网络算法。
A new method to adjust weights of BP network is proposed based on quantum particle swarm optimization.This algorithm combines concepts of particle swarm and quantum principles,which overcome the limitations both the local extreme values and bad optimization result of traditional particle swarm optimization algorithm,and get the best-optimized parameters of BP network.The new algorithm is used in simulation of three typical complex functions,results of which are analyzed and compared with that of BP algorithm and BP network optimized based on traditional particle swarm optimization algorithm.Results show the performances of the new algorithm are superior to that of other two kinds of algorithms,especially in generalization ability.
出处
《传感器与微系统》
CSCD
北大核心
2011年第7期135-136,139,共3页
Transducer and Microsystem Technologies
基金
科技部国际科技合作项目(2008DFR10530)
河北省科技厅指导性计划资助项目(072135140)
秦皇岛市科学技术研究与发展计划资助项目(201001A064)
关键词
BP网络
粒子群优化
量子论
量子粒子群优化
优化算法
权值调整
BP network
particle swarm optimization（PSO）
quantum principles
QPSO
optimization algorithm
weight adjustment
