摘要： 在分析覆盖算法的基础上提出了一种加权的覆盖领域算法,在构造神经网络覆盖领域时,充分利用了样本在球形领域间的分布位置特性及分布概率特性,做到真正意义上利用样本特征来构造覆盖神经网络,给出加权覆盖算法的构造方法及步骤,并通过实例进行了验证。

关键词: 神经网络, 覆盖算法, 权系数

Abstract: After analyzing the covering algorithm, a weighted covering algorithm is introduced in this paper. By this way, when constructing covering neighborhood of neural networks, not only the characteristics of the position of the training samples on the sphere are used, but also the distribution probability of the samples on the sphere are taken into consideration, which make full use of the learning instances to construct the covering neural network. In this paper, the method and procedures of constructing the weighting covering algorithm are given, and effect of the method is validated through experiment.

中图分类号: 

• TP18