具有部分先验信息的贝叶斯分析法在模式分类中的应用

针对模式分类中高置信度的先验概率分布难以设定的问题,提出了一种新的应用贝叶斯分析进行模式分类的方法。该方法首先求得已知先验信息和各个可能的分类决策对先验概率分布的约束集,然后通过对他们进行比较确定待分类模式的类别。实例表明该方法能够有效地缩小待分类模式的可能类别集,从而为分类系统的下一步精确分类提供便利。

Application of Bayes Analysis with Part Pre-information in Pattern Classification

To overcome the hardship of enacting the pre-probability distribution with high certainty factor,this paper proposes one novel way of applying Bayes analysis to classify pattern.This way firstly obtains pre-probability distribution sets separately restricted by the pre-information and every possible classification decisions.Then we classify the pattern by comparing them.Example shows that this way can effectively reduce the possible classes set and provide ease for classifying exactly in system.