论文部分内容阅读
建立了一种基于微粒群优化的贝叶斯网络结构学习方法,将贝叶斯网络的结构学习过程转化为对应邻接矩阵的评分寻优问题;将网络节点顺序和节点间因果关系的确定内化于评分寻优过程,避免了算法需要节点有序或事前排序的限制.建立了完整的0-1矩阵微粒群优化计算法则,在网络寻优过程中仅通过改变有向边的方向去除网络中出现的环路,以保证搜索过程中网络结构的完整性.通过ASIA网和CarStart网的数据实验证明了算法的有效性.
A Bayesian network structure learning method based on Particle Swarm Optimization (PSO) is established, which transforms the structure learning process of Bayesian network into the scoring optimization problem of the corresponding adjacency matrix. The internal sequence of nodes and the causal relationship between nodes are internalized In the process of scoring optimization, the algorithm avoids the restriction that the algorithm needs the orderly or prior order of the nodes.A complete algorithm of 0-1 matrix particle swarm optimization is established to remove the network only by changing the direction of directed edges The loop appeared to ensure the integrity of the network structure during the search process.According to the data experiment of ASIA network and CarStart network, the algorithm is proved to be effective.