面向多模态函数优化改进小生境粒子群算法.pdfVIP

2011 9 9 Journa l o f Ch inese C om puter Sy stem s V o l32 N o. 9 20 11 1, 2 2 1, 2 彭 勇 , 林 浒, 安晓荣 1 ( , 100049 2 ( , 110 17 1 Em ail: pengyong@ m ail. ustc. edu. cn : 为了保持群体多样性以增强全局搜索能力, 小生境技术在遗传算法中得到了 泛应用. 针对多模态函数优化问题, 将 小生境技术引入到粒子群算法中, 建立小生境熵作为群体多样性的量化指标, 实时考查进化过程中群体的多样性并调整进化参 数; 结合数论中的佳点理论, 提出一种在解空间使用佳点搜索的群体多样性发掘方法, 使得进化过程中群体多样性水平始终保 持在设定的阈值之上, 从而改善算法的全局搜索能力以期跳出局部最优; 在此基础上提出一种旨在找出全部全局最优解和局部 最优解的新型串行多群体小生境粒子群算法. 数值实验表明, 改进的小生境粒子群算法在求解多模态函数优化问题时具有较好 的自适应性和收敛性. 将算法应用于图像配准实验中, 使得配准参数估计误差有明显降低. : ; ; ; ; ( : T P 18 : A : ( 20 11 Improved N iche Particle Swarm Optim izer forMultim odalFunction Optim ization 1, 2 2 1, 2 PEN G Y ong , L IN H u , AN X iaorong 1 (G radua te U n iversity o f C h inese A cad em y of Sciences, B eijing 100049, C hina 2 ( Sheny an g In stitu te of C om pu ting of T echnology, C hinese A cadem y o f Sciences, Sheny an g 110 171, C h ina Abstract: Fo r m a inta in ing the popula tion d iv ersity, n iche techn ique w as w ide ly u sed in gene tic algo rithm to enhance the g lo ba l search capab ility. In th is paper, the n iche entropy w a s in tro duced to m easure the d iv ersity o f the popula tion on the ba sis o f an adaptiv e n iche identif ication approach and the evo lutionary pa ram e te rs o f the P SO a lgo r ithm can be ad ju sted adap tive ly acco rd ing to the niche en tro py o f populatio n. M oreov er, an e ff ective appro ach to explo re new schem a s in search space is designed, w h ich m akes the diversity leve l o f po pu la tion a wl ay s h ig her than the thresho ld v alue se t befo rehand dur ing the evo lutionary pro cess, so the pro po sed a lgo r ithm can o b ta in strong ab ility to search out the g loba l optmi al so lutions. A nd a no ve l sequentia lm u lt