神经网络BP算法中的误差及其改进方法研究

人工神经网络(Artificial Neural Networks)是模仿人脑结构和功能的一种信息处理系统。该方法在处理非线性问题上具有其独特的优越性,在工程应用中越来越被人们所关注。与此同时,神经网络计算的精度和收敛速度也成为人们普遍关心的问题。本文试验发现BP算法在归一化过程中存在最值误差和区间变形误差,在此基础上,本文对常规BP网络进行了改进,提高了BP网络的预测精度。

Error analysis & correction of BP network

Artificial Neural Networks is an information processing system that imitates the human brain structure and its func- tions.In dealing with non-linear issue,this method demonstrates its unique advantages,and more and more people are concerned a- bout it in engineering applications area.At the same time,the accuracy and convergence speed in Neural Networks computing have al- so been becoming a common concern.It is found that BP Neural Network Algorithm exists Max Error and Space Deformation Error in the application process.In order to control these Errors,the conventional BP Network Algorithms is improved and has enhanced its the forecast accuracy in this paper.