求解偏好多目标优化的克隆选择算法.pdfVIP

ISSN 1000-9825，CODENRUXUEW E·mail：jos@iscas．ac．ca JournalofSoftware，Vo1．21，No．1，January2010，PP．14—33 http：／w／wwjos．org．cn doi：10_3724，sP．J．1001．2010．03551 rel／Fax：+86．10 ◎byInstituteofSoftware，theChineseAcademyofSciences．Allrightsreserved． 求解偏好多目标优化的克隆选择算法 杨咚咚I2+，焦李成1,2公茂果1,2余 航 (西安电子科技大学 智能信息处理研究所，陕西 西安 710071) (智能感知与图像理解教育部重点实验室(西安电子科技大学)，陕西 西安 710071) CloneSelectionAlgorithmtoSolvePreferenceMulti-ObjectiveOptimization YANGDong．Dong，， JIAO Li．Cheng，一， GONGM ao—Guo ， YU Hang (InstituteofIntelligentInformationProcessing，XidianUniversity，Xi’an710071，China) (KeyLaboratoryofIntelligentPerceptionandImageUnderstandingofMinistry ofEducationofChina(XidianUniversity)，Xi’an710071， China) +Correspondingauthor：E·mail：ddyang@mail．xidian．edu．ca YangDD，JiaoLC，GongMG，YuH。Cloneselection algorithm to solvepreferencemulti-obJective optimization．JournalofSoftware，2010，21(1)：14—33．http：／w／ww．jos．org．cn／1000—9825／3551．htm Abstract：Thedifficultyofcurrentmulti—objectiveoptimizationcommunityliesinthelargenumberofobjectives． LackingenoughselectionpressuretowardtheParetorfont，classicalalgorithmsaregreatlyrestrained．Inthispaper， preferencerank immunememory cloneselection algorithm (PISA)isproposed tosolvetheproblem of multi—objectiveoptimizationwithalargenumberofobjectives．Thenondominatedantibodiesareproportionally clonedby theirpreferenceranks，which aredefinedbytheirpreferenceinformation．Itisbeneficialtoincrease selectionpressureandspeedupconvergencetothetruePareto·optimalfront．Solutionsusedto印prOxiIl把!ethe Pareto frontcan bereduced by preferenceinfomr ation．Becauseonly nondominated antibodiesrae selected to operate，thetimecomplexityofthealgorithm canbereduced．Besides，animmunememory populationiskeptto preserve thenondominatedantibodiesand dominance isemployed tomaintain