线性分类中基于感知器的子网分析法研究

为解决一层感知器对线性不可分矢量分类的限制，提出了一种基于一隐层感知器神经网络模型的子网分析方法。子网分析法网络构造严格精确但预处理较复杂，适合于低维矢量的分类，不会产生错分。用三维线性不可分矢量验证了这种方法的可行性。

Research on sub-network methods in linear classification based on perceptron

In order to resolve the limit of classifying the vectors of linear inseparability using perceptrons with single layer,a method of sub-network analysis is developed in this paper,which is based on the perceptrons with one hidden layer.The sub- network analysis method is strict and accurate in construction of network but more complicated in preprocessing,which is suitable for classifying low dimensional vectors and will not produce misclassifying.The feasibility of this method is proved using three di- mension vectors of linear inseparability.