一种求解多重定积分的对偶神经网络计算方法
首发时间:2017-08-25
摘要:针对多重定积分数值计算中存在效率低、精度差,甚至难以求解的问题,本文给出了一种基于对偶神经网络的多重定积分计算方法。首先给出了一种构建积分问题原函数的对偶神经网络,并从理论上证明了该方法具有以任意精度逼近任一意被积函数的原函数的能力。在此基础上,通过反复应用构建原函数的对偶神经网络方法,得到了一种能够求解任意给定积分上下限表达式的多重定积分的神经网络计算方法。算例仿真表明,本文方法是一种求解多重积分问题的高效、高精度数值计算方法。方法的优点是:(1)适用于积分限为函数的多重积分问题。(2)只需已知被积函数的有限的样本点集合,而无需知道被积函数的具体解析表达式。(3)方法的实施不受积分重数的限制。
For information in English, please click here
A dual neural network method for the solving of Multiple definite integral
Abstract:For the inefficient , less accurate and hard-to-solve problem existing in multiple definite integrals , in this study , a calculation method based on dual neural networks for the solving of multiple definite integrals is presented. Firstly, a dual neural network to construct the original function of the integral problem is presented, which theoretically proves that the method has the ability to approximate the original function of any integrand with arbitrary precision. On this basis, a neural network method for solving multiple deterministic integrals with arbitrary upper and lower bounds is obtained by applying the dual neural network method of constructing the original function repeatedly. The simulation results show that this method is an efficient and high - precision numerical method for solving multiple integral problems. The advantages of the method are: (1) The problem of multiple integration whose integral limits is functions. (2) Only a limited set of sample points of the integrand is known, without knowing the specific analytic expression of the integrand. (3) The implementation of the method is not limited by the number of integrals.
Keywords: Dual neural network Multiple definite integral Arbitrary integral domain
基金:
论文图表:
引用
No.4740858670935150****
同行评议
共计0人参与
勘误表
一种求解多重定积分的对偶神经网络计算方法
评论
全部评论0/1000