人工智能中的贝叶斯方法  

Bayesian analysis in artificial intelligence

作　　者：程和祥 聂炜昌 CHENG Hexiang;NIE Weichang(School of Administrative Law (School of Supervision Law), Southwest University of Political Science and Law, Chongqing 401120, China;Law School of Central China Normal University, Wuhan 430079, China)

机构地区：[1]西南政法大学行政法学院(监察法学院),重庆401120 [2]华中师范大学法学院,湖北武汉430079

出　　处：《重庆理工大学学报（社会科学）》2020年第5期17-23,共7页Journal of Chongqing University of Technology（Social Science）

基　　金：教育部人文社会科学研究规划基金项目“可能与必然的概念史研究”(19YJC720029);西南政法大学人工智能研究院重点课题“人工智能在司法裁判中的应用及其限度”(2018-RGZN-JS-ZD-05)。

摘　　要：贝叶斯方法有狭义和广义之分,狭义的贝叶斯方法是以贝叶斯定理为核心的概率归纳逻辑,广义的贝叶斯方法是一种科学哲学研究纲领,它兼顾了背景知识和理性分析的重要性。朱迪·珀尔在人工智能研究中引入贝叶斯网络方法,使人工智能研究取得了突破性进展。根据贝叶斯方法逐步完善的程度,人工智能中的贝叶斯方法可以分为贝叶斯规则、贝叶斯概率推理和贝叶斯网络3个阶段。Bayesian analysis can be defined in a narrow sense and a broad sense.In a narrow sense,Bayesian analysis is inductive probability logic with Bayes’theorem as its core.Broadly speaking,Bayesian analysis is a research program of the science and philosophy that possesses the importance of background and rational analysis.Judea Pearl has introduced the Bayesian networks method in artificial intelligence research,which made breakthroughs in the artificial intelligence research field.According to the degree of perfection of Bayesian analysis,the application of Bayesian analysis in artificial intelligence field can be divided into three stages:Bayesian rule,Bayesian probability reasoning,and Bayesian networks.

关 键 词：贝叶斯规则 贝叶斯概率推理 贝叶斯网络 

分 类 号：B81[哲学宗教—逻辑学]