一类反馈神经网络对非线性连续系统近似能力的证明

摘要

Starting from the universal approximation theorem of multilayered feed forward neural networks, this paper proves that the finite time trajectory of nonlinear continuous-time system with input can be approximated by the state vector of the output units of a class of recurrent neural networks.%本文从多层前馈神经网络的一般近似定理出发，证明了带有输入的非线性连续系统在有限时间段内的输出轨迹可以被一类反馈神经网络输出神经元的状态向量近似到任何程度。