一种基于泛函网络求数值积分方法研究  

Numerical Integration Method Study Based on Function Network

作　　者：韦修喜[1] 周永权[2] 蓝晓玲[1] 

机构地区：[1]广西大学计算机与电子信息学院,南宁530004 [2]广西民族大学数学与计算机科学学院,南宁530006

出　　处：《计算机科学》2009年第4期224-226,238,共4页Computer Science

基　　金：国家自然科学基金(60461001);广西自然科学基金(0542048;0832082);国家民委科研项目(08GX01);广西研究生教育创新计划资助项目(T32084)资助

摘　　要：提出了一种基于泛函网络求数值积分新方法,给出了一种泛函网络模型及学习算法,并将该模型用于求任意函数的数值积分,理论上证明了泛函网络用于逼近数值积分定理。最后通过5个数值积分算例,并与传统计算方法作了比较分析,仿真结果表明,提出的数值积分方法精度高,适应性强,且不需要确定被积函数的原函数,因此该方法在工程技术中有较大的应用价值。A new method based on functional network was proposed to solve numerical integration methods. A functional network model for the numerical approximation and learning algorithm were given, and its model was used to sovle numerical integration of arbitrary functions. The theorem of numerical integration solution based on functional network theorem was presented and proved. Finally, five experimental results show that the new measure has higher precision and strong adaptability, and primitive functions of integrand need not to be confirmed. Therefore, this new measure has application values in many engineering areas.

关 键 词：泛函网络 基函数 学习算法 函数逼近 数值积分 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]