PAS理论与应用.doc

河北工业大学
硕士学位论文
PAS理论与应用
姓名:叶建伟
申请学位级别:硕士
专业:通信与信息系统
指导教师:伍萍辉;曾成
20071101
河北工业大学硕士学位论文
PAS 理论与应用
摘要
在信息融合技术中,受传感器自身局限和外部环境的影响,最初得到的数据往往是不
准确的,如何将这些不准确的数据进行有效的融合,推出基本正确的结论,是不确定推理
理论要研究的内容,而概率论据系统作为一种简单有效的不确定推理理论,目前已经广泛
应用于许多实用领域,因此,对概率论据系统进行系统研究具有重要意义。本文所作的主
要工作如下:
首先描述概率论据系统的理论基础,阐述了准支持集合、支持集合以及支持度等基本
概念,分析了准支持集合的传统计算方法所存在的缺陷,提出一种较优的矩阵边际化算法,
并用实例验证其有效性。
其次,通过划分算法把概率论据系统和证据组合理论联系起来,使得对支持度的计算
可以转化为多个证据的组合。并详细介绍了与证据组合相关的熔合算法与连接树理论。为
得到相对最优的变量消去顺序,提出一种启发性算法,此算法可有效弥补已有启发性算法
的不足,与其它算法一起使用时,具有较好效果。
第三,阐述了基于连接树理论的内外传播局部计算方法,此方法在处理中型问题时尤
为适用,在此基础上初步探讨了证据组合并行计算的可行性。
最后,叙述了一种专为概率论据系统开发的软件语言-ABEL,并通过两个实例演示
使用ABEL 的简单性和有效性。
关键词:概率论据系统,证据组合,启发性算法,局部计算
i
PAS 理论与应用
PAS THEORY AND ITS APPLICATIONS
ABSTRACT
Constrained by the sensor itself and the circumstance, the original data isn’t always exactly
right in the information fusion. How to fuse these data in an effective way to get the basically
accurate results, is the most important task of urate reasoning theory. As a simple and
effective theory, Probabilistic Argumentation System(PAS)has been implied in many practical
fields. So, it is necessary to research this theory systemically. The main tasks of this dissertation
are as follows:
Firstly, the basic theory and the fundamental concepts of the PAS, such as the quasi
supporting set,supporting set and the supporting degree, is des After analyzing the default of
the traditional algorithm which based on the quasi supporting set, a new algorithm, superior
than the one mentioned above, named as the Matrix Marginal algorithm is proposed and the
affectivity is proved by a concrete example.
Secondly, the PAS and the Dempster-Shafer Theory(DST) is connected by a partition
algorithm, so that putation of the supporting degree is transformed to position
of many evidences. There are detailed presentation of fusion algorithm and join tree theory
which are correlative with DST. In order to get the relatively optimal ranking of the elimination
of varia