摘要
本文讨论了模型管理中不确定性的表达、传递、证据合成以及问题求解过程,提出了一种基于非单调逻辑的模型管理方法:将模型结构形式的不确定性表示为由建模者或领域专家对问题结构中未知或随机情形所作假设集支持的可能性命题;模型之间不确定性关系的管理通过对假设环境的真值(一致性)保持和信度调整过程来实现,其依据是在问题求解过程中出现的冲突情形或者是由决策人提供的有关命题或次判断。
This paper presents a nonmonotonic logic-based approach to the problem of model management in uncertain environments. The proposed algorithm for uncertainty analysis provides a mechanism for uncertainty representation, propagation, evidence combination and problem solving under uncertainties. Furthermore, uncertainties about model structures, i.e., the probabilities of possible model structures, are represented by probability propositions supported by sets of ATMS's assumptions about uncertain situations. Hence, the management of uncertain causal link between models is realized in the processes of truth maintenance, belief assignment and adjustment about assumptions' environment, wherein, conflict situations presented in problem solving processes and related ATMS justifications about the results of problem solving are taken as new evidences for truth maintenance and the adjustment of degrees of belief.
出处
《自动化学报》
EI
CSCD
北大核心
1992年第4期414-420,共7页
Acta Automatica Sinica
关键词
模型管理
非单调逻辑
Model management
nonmonotonic logic
assumptions' environment
uncertainty
justifications
