论文部分内容阅读
神经网络是一种普遍使用的分类方法.当类别数目较大时,神经网络结构复杂、训练时间激增、分类性能下降.针对这些问题,基于N分类问题的两分类方法和树型分类器结构,对两分类子网络集进行排序,文中给出了一种大类别分类的神经网络阵列结构和快速搜索方法并重点分析了网络阵列的分类性能.理论分析表明,使用网络阵列方法可降低平均分类错误率.该方法还使得网络结构简单灵活,易于扩充,网络的训练时间缩短.仿真实验表明,该方法对类别数大且分类困难的复杂分类问题的分类效果良好.
Neural network is a commonly used classification method. When the number of categories is large, the structure of neural network is complex, the training time increases rapidly and the classification performance decreases. Aiming at these problems, the two-class classification based on the N-classifier and the tree classifier are used to sort the two-class sub-network sets. In this paper, a large-scale classification neural network array structure and fast search method are given, Network array classification performance. Theoretical analysis shows that using the network array method can reduce the average classification error rate. The method also makes the network structure simple and flexible, easy to expand, the network training time is shortened. Simulation results show that the proposed method is effective in the classification of complex classification problems with large number of classes and difficult classification.