摘要
BP算法存在收敛性慢、逼近精度差等缺点。因此,在实际应用中往往需要对BP算法进行相应的改进。利用主元分析法(PCA)对数据样本预先进行降维处理,消除数据间的强耦合性,减少模型的复杂性,然后再作为BP的输入样本从而提高模型的解算速度。在此分析的基础上利用Visual Basic采用模块化的设计方法实现对PCA-BP算法的编程,使学习速率能够进行相应自调整和优化,以此来提高BP网络的泛化推广能力,并能够满足设定的误差精度从而达到现场实际运用需要的目的。
Traditional BP algorithm has slow convergence and weak approximability. Therefore,must improve BP algorithm accordingly. Introduces the method that can reduce dimensions of datum by PCA and eliminate the strong coupling metric,reducing complexity of model,the result is the input stylebook of BP in order to improye the speed of calculation.Depending on the base of analysis,can realize PCA - BP algorithm program by Visual Basie,accelerating and optimizating the rate of algorithm learning,in order to improve the generalization and satisfy given error precision reaching to practical application ability of BP network.
出处
《计算机技术与发展》
2008年第12期98-101,105,共5页
Computer Technology and Development
基金
安徽省高校青年教师科研资助计划项目(2005jql232)
