爬山法与模式蚁群法混合的贝叶斯优化算法  

Hill-climbing and pattern ant colony hybridBayesian optimizat ion algorithm

作　　者：胡云安[1] 刘振[1] 宋瑞华[2] 史建国[3] 

机构地区：[1]海军航空工程学院控制工程系,山东烟台264001 [2]海军航空工程学院训练部,山东烟台264001 [3]海军航空工程学院兵器科学与技术系,山东烟台264001

出　　处：《华中科技大学学报（自然科学版）》2013年第5期90-95,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61174031;60674090)

摘　　要：针对传统贝叶斯优化算法进化效率低及收敛速度慢的情况,提出一种新型混合贝叶斯优化算法.该算法利用适应度遗传及个体的局部搜索方法,使种群个体趋向于全局最优解,提高了进化效率.为提高贝叶斯优化算法中贝叶斯网络结构学习的效率,提出一种爬山法和模式蚁群算法相结合的网络结构学习方法,同时对新型贝叶斯优化算法的收敛性进行了分析.利用典型的函数对提出的新型混合贝叶斯优化算法进行了仿真分析,证明了所提出的方法可以有效地加快算法的收敛速度和收敛精度.将该算法应用于目标分配问题中,仿真证明了所提算法的有效性和优越性.To overcome the low evolution efficiency and long conve rgence time in the traditional Bayesian optimization algorithm（BOA）,a new hybr id BOA was proposed.The fitness inherence and individual local search were inco rporated in the new algorithm to enhance the evolution efficiency.In order to p romote the construction speed of Bayesian network structure,hill-climbing and pattern ant colony was used to learn the structure of Bayesian network.The conv ergence of new hybrid BOA was also analyzed in the paper.The simulation results of benchmark functions show that the new hybrid algorithm performs well than be fore in promoting the precision and reducing the time complexity.The new algori thm was used to target allocation problem.The validity and superiority of the a lgorithm were also proved by the simulation results.

关 键 词：贝叶斯优化算法 爬山法 局部搜索 模式 蚁群算法 目标分配 

分 类 号：TP15[自动化与计算机技术—控制理论与控制工程]