基于诱导向量的CC4神经网络行为研究

Behavior of CC4 Neural Network Based on Deriving Vector

角分类前向神经网络CC 4可以快速对文本数据进行分类处理。本文在定义二值向量的诱导向量的基础上,给出CC 4神经网络隐层、输出层的权矩阵构造方法的诱导向量分析,并给出CC 4神经网络隐层输出的基本原理、基于泛化距离的隐层权矩阵构造方法的几何解释,以及输出层权矩阵构造的约束条件;揭示了角分类神经网络学习、工作的基本原理。本文为基于实向量输入的快速角分类神经网络的设计提供了借鉴及必要的理论基础。

A feed forward neural network （FFNN） for the corner classification 4 （CC4） can instantaneously classify text data. With the definition of the deriving vector for a binary vector, the deriving vector based analyses for constructions of weight matrixes of the CC4 hidden and output layer are given. The principle for outputs of the CC4 hidden layer, and the geometrical interpretation for the construction of the CC4 hidden weight matrix, and constrains for the construction of the CC4 output weight matrix are also presented. Research conclusions show that the fundamental principle of the learning and the running of FFNN for corner classification provide some references and the academic foundation for designing the new FFNN for corner classification with the real vector as inputs.