本文首先对饱和模式线性神经网络模型(LSSM)进行分析,指出其实际上是一种沿内部和边界面依次搜索局部极小值点进行存储的联想记忆模型,进而证明了基于约束区域的神经网络与LSSM具有相同的联想记忆特性,它由普通的常微分方程描述,更利于模拟和硬件实现,计算机模拟说明了结论的正确性。 This paper first analyzes the saturated mode linear neural network model (LSSM), points out that it is actually a kind of associative memory model which searches for the local minimum points along the interior and the boundary, and then proves that the neural network based on the constrained region It has the same associative memory characteristic with LSSM. It is described by ordinary ODE, which is more conducive to both simulation and hardware implementation. Computer simulation shows the correctness of the conclusion.