基于遗传算法优化神经网络的技术研究  

The Research on Technology Based on Genetic Algorithm and Neural Network

作　　者：高宪军[1] 张杰[1] 洪俞[1] 程国晓[1] 

机构地区：[1]空军航空大学航空电子工程系,吉林长春130022

出　　处：《装备制造技术》2010年第2期8-10,共3页Equipment Manufacturing Technology

摘　　要：为克服传统BP算法的收敛速度慢和容易陷入局部极小值的缺点,提出了一种基于遗传算法优化的BP神经网络智能诊断技术:采用基于实数编码的遗传算法优化神经网络权值和阈值,代替了原BP网络随机设定的初始权值和阈值,然后再用BP算法对网络进行精确搜索。实验仿真结果表明,用此法优化过的神经网络的训练步数明显减少,诊断准确率达91.7%,泛化能力也得到提高。In order to overcome the shortcomings of traditional BP, such as low convergence speed, easy to fall into the local minimum points, an intelligent diagnosis method which is based on genetic algorithm optimized BP neural networks is presented. An improved genetic algorithm based on real-coding is adopted,the primitive weight value and threshold devised random are replaced ,then net is searched precisely by BP.The simulation results indicate that the genetic algorithm optimized Bp neural networks train＇s epoch decrease, the rate of precision could be attained by 91.7%,and generalization ability has been enhanced.

关 键 词：神经网络 遗传算法 技术 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]