一类输入元素非线性连接的反馈式神经网络及其联想能力分析

本文研究了输入元素非线性连的妆的反馈式神经网络，文中以二阶非线性连接为例给多拓扑结构，导出了能够实现模式平移不变识别的学习方法，并借助于等权类的概念把不变识别条件建造于网络权结构之中，同时降低了网络连接复杂度。

On the Associative Ability of a Feedback Neural Network with Nonlinearly Connected Input Elements

In this paper,a feedback neural network,of which input elements are nonlinearly connected, is studied. The topology of its second order case is shown,the learning rules which make shift pattern recognition possible are derived. With the help of the idea of the equivalent weight subset,not only the shift pattern invariance conditions are coded in weight matrix but the nonlinearly connected network is made much simpler.