摘要
为了求得在VC环境下几种改进BP算法中整体性能最优的一种BP算法,介绍了三种改进BP算法,并在一个用VC开发的仿真平台上对传统BP算法和三种改进算法的稳定性、收敛速度等性能进行了对比和分析。结果表明,基于权值变化量的迭代算法收敛速度最快,综合性能较优。仿真结果对实际工程应用中选择合适的BP训练法,有一定的指导意义。
In order to find BP algorithm that has the best synthesis performance among several fast BP algorithm by using VC IDE,this paper presents three fast BP algorithms,and the capablilities are contrasted via the emulator realized by VC.It is demonstrated that the fast BP algorithm based on iterative weight-change extent has the quickest convergence rate and better performance.The result has positive function in choosing the right BP algorithm applied to engineering application.
出处
《电脑开发与应用》
2005年第7期45-46,49,共3页
Computer Development & Applications
关键词
神经网络
权值调整
快速BP算法
性能
neural network,weights adjustment,fast BP algorithm,performance
