基于讨论机制的头脑风暴优化算法  

Discussion mechanism based brain storm optimization algorithm

作　　者：杨玉婷[1,2] 史玉回[3] 夏顺仁[1,2] 

机构地区：[1]浙江大学生物医学工程教育部重点实验室,浙江杭州310027 [2]浙江省心脑血管检测技术与药效评价重点实验室,浙江杭州310027 [3]西交利物浦大学电子与电气工程系,江苏苏州215123

出　　处：《浙江大学学报（工学版）》2013年第10期1705-1711,1746,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60772092;81101903);国家"十一五"科技支撑计划资助项目(2012BAI10B04)

摘　　要：为了克服头脑风暴优化(BSO)算法易陷入局部最优导致早熟收敛的问题,提出新型的基于讨论机制的头脑风暴优化(DMBSO)算法.该算法运用组内讨论和组间讨论这一新机制取代BSO算法中的个体更新过程,分别控制算法的全局搜索和局部搜索能力.通过线性递减和线性递增方式调整组间讨论和组内讨论次数,使算法搜索初期加强全局搜索能力,搜索后期加强局部细致搜索能力,有效地防止早熟问题.对6个经典测试函数(BFs)的10维、20维、30维问题分别进行测试来评估DMBSO的效果.结果表明,DMBSO算法与BSO算法和经典的粒子群(PSO)算法相比,可以有效地避免陷入局部最优,稳定地找到更好的最优值,而且随着问题维度的增加,DMBSO表现出更强的鲁棒性.A discussion mechanism based brain storm optimization （DMBSO） algorithm was proposed in order to solve the problem that brain storm optimization （BSO） algorithm is likely to stagnate in the local optima and result in premature convergence. DMBSO used a new mechanism with inter-group discussion and intra-group discussion to replace the process of individual updating in the original BSO algorithm in order to respectively govern the ability of global search and local search. The ability of global search was enhanced at the beginning by linearly decreasing times of inter-group discussion and increasing times of intra-group discussion, while fine search was enhanced in the end to prevent premature convergence. Empirical studies were conducted to evaluate the performances of the DMBSO algorithm for the 10D, 20D, 30D problems of six popular benchmark functions （BFs）. Experimental results demonstrate that the DMBSO algorithm can avoid being stagnated in the local optima, more effectively and steadily find the better results than the original BSO algorithm and standard particle swarm optimization （PSO） algorithm, and show stronger robustness with the increasing of BFs＇ dimension.

关 键 词：群体智能优化算法 头脑风暴优化算法 讨论机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]