径向基函数在文本分类中的应用

利用RBF网络进行文本自动分类,对训练集进行聚类,使得每个簇内部的相似性尽可能高,而簇之间的相似性尽可能低.为每一个簇的中心定义相应的径向基函数,再对由这些径向基函数构成的两层神经网络进行训练.因考虑到了不同簇之间的差异性,因此很好解决＂兼类＂问题.

Application of RBF Network on Text Categorization

This paper introduces the method of using RBF network for automated text categorization.The main idea is that we can get some clusters which have high similarity inside one cluster and low similarity between pairs of clusters and define a Radial Basis Function at center of each cluster and learn a two-layer neural network which consists of these RBFs.Since we have considered the discrimination between different clusters in a class,the method can tackle the problem of multi-classes excellently.