【摘要】 本文利用多层前向神经网络的函数逼近能力,推导出一阶微分方程组边值问题的神经网络解法,为一阶微分方程组的近似解析解的求取提供了便于工程应用的新方法.更多还原

【Abstract】 Based on the theory that a hidden multi-feedforward neural network can approximate a square integral function with an arbitrary accuracy, a method to solve the first order diiferential equations using neural networks is proposed. The approximated solution to the first order differential equations can be achieved which is useful for practical applications.更多还原