欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
贝叶斯网络结构学习对贝叶斯网络解决实际问题至关重要．基于评分与搜索的方法是目前比较常用的结构学习方法，但该类方法中结构搜索空间的大小随结点个数增加而指数增长，因此一般采用启发式搜索策略，有些方法还需要结点次序．在基于结点次序的最大相关．最小冗余贪婪贝叶斯网络结构学习算法中，由于是随机产生初始结点的次序，这增大了结果的不确定性．本文提出一种生成优化结点初始次序的方法，在得到基本有序的结点初始次序后，再结合近邻交换算子进行迭代搜索，能够在较短的时间内得到更加正确的贝叶斯网络结构．实验结果表明了该方法的有效性．
英文摘要：
Bayesian network structure learning is critically important when using Bayesian network to solve practical problem. Al- though the method at present that based on Search ＆ Score is more and more popular, the search space of structure increased exponen- tially with the number of nodes, so usually adopt heuristic search methods and some algorithm require node ordering. When learning structure of Bayesian network by Ordering-based Max-Relevance and Min-Redundaney Greedy algorithm, node＇s initial order pro- duced at random will increase the uncertainty of the result greatly. In this paper, we propose a new method that can find a basic order- ly order. This method combined with neighbor-swap operator to search iteratively, can obtain a better result in a shorter time. Experi- mental results show that the new approach is correct and effective.
同期刊论文项目
同项目期刊论文
期刊信息
位置：