神经网络定性分析

A Method of Qualitative Analysis of Neural Networks

利用大系统理论和M-矩阵的特性,提出了一种适合于Hopfield神经网络平衡点定性分析的新方法,给出了系统稳定和不稳定的判定定理。把神经网络的不同部分看成是子系统,使稳定性判定过程中的计算量大大减少。对非线形互联结构强度的约束不仅可以为直线,也可以是一些适当的函数,扩大了适用范围。

A new method for qualitative analysis of equilibrium points of Hopfield neural networks is given by using large-scale system theory and the quality of Minkowski matrix. Two qualitative theorems for neural networks are given both stability and instability. The amount of computation is reduced greatly viewed the different part of networks as subsystems. The restraint of intensity of the nonlinear interconnecting structure is not only a line, but also some suitable function.