基于遗传算法的贝叶斯网络结构学习综述  

A review of Bayesian network structure learning based on Genetic Algorithm

作　　者：朱宇 王慧玲 郑锦波 綦小龙 ZHU Yu;WANG Huiling;ZHENG Jinbo;QI Xiaolong(School of network security and information technology,Yili Normal University,Yining 835000,China)

机构地区：[1]伊犁师范大学网络安全与信息技术学院,新疆伊宁835000

出　　处：《激光杂志》2023年第4期32-39,共8页Laser Journal

基　　金：新疆维吾尔自治区自然科学基金项目(No.2021D01C467、No.2022D01C337);新疆维吾尔自治区高校科研项目(No.XJEDU2020Y036);伊犁师范大学博士科研启动项目(No.2020YSBS007)。

摘　　要：贝叶斯网络采用图模型描述变量之间的依赖关系,因其结构清晰,具有突出的决策机制和学习机制,故拥有优秀的推理能力。在各类研究方法中,遗传算法能够有效地解决复杂的优化问题,以其普适性好、鲁棒性强、便于并行执行、高效便捷等显著特点,在贝叶斯网络结构的学习研究过程中发挥着非常重要的作用。从初始种群、遗传操作算子设计两个层面对近年基于遗传算法的因果结构学习改进方法进行了调研分析并指出了该技术路线进一步的研究方向。Bayesian network uses a graph model to describe the dependencies between variables.Because of its clear structure and outstanding decision-making mechanism and learning mechanism,it has excellent reasoning ability.Among various research methods,genetic algorithm can effectively solve complex optimization problems.With its remarkable features such as good universality,strong robustness,easy parallel execution,high efficiency and convenience,it is widely used in the learning and research process of Bayesian network structure.plays a very important role.This paper investigates and analyzes the improvement method of causal structure learning based on genetic algorithm in recent years from the two levels of initial population and genetic operation operator design,and points out the further research direction of this technical route.

关 键 词：贝叶斯网络结构学习 搜索算法 遗传算法 

分 类 号：TN249[电子电信—物理电子学]