According to the soundness and completeness of information in databases, the expressive form and the semantics of incomplete information are discussed in this paper. On the basis of the discussion, the current studies...According to the soundness and completeness of information in databases, the expressive form and the semantics of incomplete information are discussed in this paper. On the basis of the discussion, the current studies on incomplete data in relational databases are reviewed. In order to represent stochastic uncertainty in most general sense in the real world, probabilistic data are introduced into relational databases. An extended relational data model is presented to express and manipulate probabilistic data and the operations in relational algebra based on the extended model are defined in this paper.展开更多
现有的关系学习研究都是基于完备数据进行的,而现实问题中,数据通常是不完备的.提出一种从不完备关系数据中学习概率关系模型(probabilistic relational models,简称PRMs)的方法——MLTEC(maximum likelihood tree and evolutionary com...现有的关系学习研究都是基于完备数据进行的,而现实问题中,数据通常是不完备的.提出一种从不完备关系数据中学习概率关系模型(probabilistic relational models,简称PRMs)的方法——MLTEC(maximum likelihood tree and evolutionary computing method).首先,随机填充不完备关系数据得到完备关系数据.然后从每个随机填充后的数据样本中分别生成最大似然树并作为初始PRM网络,再利用进化过程中最好的网络结构反复修正不完备数据集,最后得到概率关系模型.实验结果显示,MLTEC方法能够从不完备关系数据中学习到较好的概率关系模型.展开更多
文摘According to the soundness and completeness of information in databases, the expressive form and the semantics of incomplete information are discussed in this paper. On the basis of the discussion, the current studies on incomplete data in relational databases are reviewed. In order to represent stochastic uncertainty in most general sense in the real world, probabilistic data are introduced into relational databases. An extended relational data model is presented to express and manipulate probabilistic data and the operations in relational algebra based on the extended model are defined in this paper.
基金Supported by the National Natural Science Foundation of China under Grant Nos.60635030 60473046 (国家自然科学基金)+1 种基金the China Postdoctoral Science Foundation under Grant No.20060390921 (中国博士后科学基金)the Jiangsu Planned Projects for Postdoctoral Research Funds of China under Grant No.0601017B (江苏省博士后科研资助计划)
文摘现有的关系学习研究都是基于完备数据进行的,而现实问题中,数据通常是不完备的.提出一种从不完备关系数据中学习概率关系模型(probabilistic relational models,简称PRMs)的方法——MLTEC(maximum likelihood tree and evolutionary computing method).首先,随机填充不完备关系数据得到完备关系数据.然后从每个随机填充后的数据样本中分别生成最大似然树并作为初始PRM网络,再利用进化过程中最好的网络结构反复修正不完备数据集,最后得到概率关系模型.实验结果显示,MLTEC方法能够从不完备关系数据中学习到较好的概率关系模型.