基于幂基函数变步长神经网络求解数值积分  

SOLVING NUMERICAL INTEGRATION BASED ON POWER BASIS VARIABLE STEP NEURAL NETWORK

作　　者：李哲[1] 王冬冬[1] 

机构地区：[1]青岛滨海学院大专理科基础学院,山东青岛266555

出　　处：《计算机应用与软件》2011年第9期123-125,共3页Computer Applications and Software

基　　金：广西自然科学基金(0832082);青岛滨海学院科技立项(2009K09)

摘　　要：提出一种基于幂基函数变步长神经网络算法求解数值积分的新方法,证明了该算法的收敛性以及数值积分的求解定理及推论。通过典型数值积分算例,计算机仿真实验表明,提出的基于幂基函数变步长神经网络积分算法相比传统的数值积分方法,具有计算精度高、收敛速度快、算法稳定等特点。A new approach for solving numerical integration problem based on power basis variable step neural network algorithm is put forward,and the convergence of this algorithm,the theorems of solution and the deductions of numerical integration are proved.Computer simulation experiments show through examples of typical numerical integration that compared with traditional numerical integration method,the approach presented in the paper has the characteristics of higher convergence rate,higher accuracy and higher stability.

关 键 词：神经网络 变步长 幂基函数 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]