摘要
针对启发式算法应用于贝叶斯网络推理学习易陷入局部最优和寻优效率低的问题,提出一种基于混合粒子群-差分法的贝叶斯网络推理算法.该算法利用自适应的反向学习策略增加初始种群的多样性,将差分变异算子引入离散粒子群算法,提出自适应概率分层搜索策略平衡局部搜索与全局搜索,并根据levy飞行机制建立自适应的变异策略避免算法陷入局部最优.由算法的收敛性分析可知,通过迭代搜索可以找到贝叶斯网络的最大可能解释.实验结果表明与其他算法相比收敛精度与寻优效率均有提升.
Aiming at the problem that heuristic algorithm applied to Bayesian network inference learning is easy to fall into local optimum and inefficient in optimization,a new Bayesian network inference learning algorithm based on hybrid particle swarm optimization-difference algorithm(HDPSO-DE)is proposed.The adaptive reverse learning strategy is adopted to increase the diversity of initial population,The differential mutation operator is introduced into particle swarm optimization algorithm,and an adaptive probability hierarchy search strategy is proposed to balance local search and global search.Meanwhile,an adaptive mutation strategy is established according to Levy flight mechanism to avoid the local optimum.The convergence analysis of the proposed algorithm demonstrates that most probable explanation can be found through the iterative search.The experimental results demonstrate that the convergence accuracy and optimization efficiency are improved compared with other algorithms.
作者
范瑞星
刘浩然
张力悦
苏昭玉
刘彬
FAN Rui-xing;LIU Hao-ran;ZHANG Li-yue;SU Zhao-yu;LIU Bin(School of Electrical Engineering,Yanshan University,Qinhuangdao 066004,China;School of Information Science and Engineering,Yanshan University,Qinhuangdao 066004,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第6期1156-1162,共7页
Journal of Chinese Computer Systems
基金
河北省人才工程培养资助项目(A201903005)资助
河北省自然科学基金项目(F2019203320)资助
国家自然科学基金项目(51641609)资助.
关键词
贝叶斯网络推理算法
粒子群算法
差分算法
levy飞行机制
Bayesian network inference learning
particle swarm optimization
differential evolution
levy flight