基于贝叶斯最优分类器的多源模糊信息融合方法(英文)

• 作 者: 苏宏升
• 机 构: School of Automatic and Electrical Engineering Lanzhou Jiaotong University,,Lanzhou 730070,P.R.China
• 出 处: 自动化学报
• 发表日期: 2008年03期
• 关键词: classifier Bayesian vague 分类器 retain reasoning 信息融合方法 directions 贝叶斯 prior

To make conventional Bayesian optimal classifier possess the abilities of disposing fuzzy information and realizing the automation of reasoning process,a new Bayesian optimal classifier is proposed with fuzzy information embedded.It can not only dispose fuzzy information effectively,but also retain learning properties of Bayesian optimal classifier.In addition,according to the evolution of fuzzy set theory,vague set is also imbedded into it to generate vague Bayesian optimal classifier.It can simultaneously simulate the twofold characteristics of fuzzy information from the positive and reverse directions.Further,a set pair Bayesian optimal classifier is also proposed considering the threefold characteristics of fuzzy information from the positive,reverse,and indeterminate sides.In the end,a knowledge-based artificial neural network(KBANN)is presented to realize automatic reasoning of Bayesian optimal classifier.It not only reduces the computational cost of Bayesian optimal classifier but also improves its classification learning quality.

To make a conventional Bayesian optimal classifier with abilities of disposing fuzzy information and realizing the automation of reasoning process, a new Bayesian optimal classifier is proposed with fuzzy information embedded. It can not only dispose fuzzy information effectively, but also retain learning properties of Bayesian optimal classifier.In addition, according to the evolution of fuzzy set theory, vague set is also imbedded into it to generate vague Bayesian optimal classifier. It can simultaneously simulate the twofold characteristics of fuzzy information from the positive and reverse directions. Focus, a set pair Bayesian optimal classifier is also proposed considering the threefold characteristics of fuzzy information from the positive, reverse, and indeterminate sides.In the end, a knowledge-based artificial neural network (KBANN) is presented to realize automatic reasoning of Bayesian optimal classifier. It not only reduces the computational cost of Bayesian optimal classifier b ut also improves its classification learning quality.