一种信度网推理的并行计算模型  

A Parallel Computing Model on Belief Network Reasoning

作　　者：张聪[1] 沈一栋[2] 刘启元 

机构地区：[1]重庆大学计算机学院 [2]中国科学院软件研究所计算机科学重点实验室,北京100080 [3]重庆市电子商务有限公司,重庆400012

出　　处：《计算机工程与应用》2004年第29期40-43,72,共5页Computer Engineering and Applications

基　　金：国家自然科学基金项目(编号:69883009;60373052);跨世纪优秀人才培养计划基金项目;重庆市科技攻关项目:面向工业应用的智能开发平台及系统研究资助

摘　　要：基于概率知识表达的信度网已成为人工智能中非确定知识表达和推理的研究热点。推理算法是信度网学习和应用的基础。该文提出了一种基于经典Polytree算法的推理计算模型。该模型表达清楚,计算过程容易控制,并能够简单地映射到并行机结构上。该文首先介绍了模型在单联通网络下的计算步骤,然后将模型引入到多联通网络上。Belief network based on probabilistic knowledge representation has become the research point in non-deterministic knowledge representation and inference in AI.The inference algorithm is the basis of learning and application in belief network.This paper proposed a new calculation model based on the classic algorithm:Polytree Propagation algorithm.The concept in this model is very clear,and the calculation process is easy of control and can be mapped to parallel structures of current computers easily.This paper first introduces the calculation steps of the model in singly connected belief networks and the way to implement parallel computing,then introduces the utilization of such model in multiply connected belief networks.

关 键 词：人工智能 信度网 概率推理 知识表达 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]