论文部分内容阅读
在对高维大数量模式样本进行分类时,一般都要用逐步聚类方法多次完成。首先要尽量获取和利用先验知识进行初始划分,找出凝聚中心;再逐步调整进行细分类。本文提出一种新的聚类方法——(0,1)模型阵法。它可简单快速地完成高维大数量模式样本的粗分类,对提高后续的细调整分类提供一个良好的初始条件。在模式样本可分性较好或精度要求不高情况下,此方法是一种快速一次完成聚类法,不必再作后续精调。
In the classification of high-dimensional large number of pattern samples, it is generally necessary to step through the clustering method to complete multiple times. First of all, we should try to obtain and use the prior knowledge of the initial division, to find the center of aggregation; and then gradually adjust the classification. This paper presents a new clustering method - (0,1) model matrix method. It can finish the rough classification of high-dimensional and large-volume pattern samples easily and quickly, and provides a good initial condition for improving the subsequent fine adjustment classification. This method is a fast and complete clustering method without any further refinement because of the good separability of the pattern samples or the low accuracy requirement.