多维正态分布分段禁忌优化算法的参数优化.pdfVIP

ComputerEngineeringandApplications计算机工程与应用 2015，51(6) 59 林 多维正态分布分段)qr ·t 、优化算法的参数优化 冀荣华 ，李 想 ，高万林 ，王枫辰 ，祁力钧 JIRonghua，LIXiang，GAOW anlin，WANGFengchen，QILijun2 1．中国农业大学 信息与电气工程学院，北京 100083 2．中国农业大学 工学院，北京 100083 1．CollegeofInformationandElectricalEngineering，ChinaAgriculturalUniversity,Beijing100083，China 2．CollegeofEngineering，ChinaAgriculturalUniversity,Beijing100083，China JIRonghua，LIXiang，GAO W anlin，eta1．Parametersoptimization ofstaged continuousTabu search algorithm basedonmultidimensionalnormaldistribution．ComputerEngineeringandApplications，2015，51(6)：59—63． Abstraet：Therearemanysho~agesfortheoptimizationofmultidimensionalcontinuousfunction．suchasthelOW conver— gencespeedandthepooraccuracy．Inordertoimprovetheglobaloptimizationperfomr anceofmultidimensionalcontinuoUS ufnction，thispaperproposestheparametersoptimizationoftheModifiedStagedContinuousTabuSearchalgorithm(MSCTS) basedonmultidimensionalnormaldistribution．MSCTSisbasedonStagedContinuousTabuSearchalgorithm (SCTS)，which includesthreestagesandhasdifferentstartingpoints，selectionrangeofneighborhoodsandsearchingstepsizeatdifferent stages．Togetthebestaccuracyandhighconvergencespeed，themultidimensionalnormaldistributionfunctionisusedto givethedirectionofneighborhoodsselection．Theresultsofexperimentsshow thatthereductionratioofsolutionspacein secondstagehasgreatimpactontheaccuracyandconvergencespeed，andthebestreductionratiois0．5．Atthesametime， toimprovetheaccuracy,thestep sizeisshrinkingatthethirdstage．Theresultsofexperimentsshow thebestshrinkingratio is0．1．ComparedwithSCTSalgorithm，especiallytosolvemultidimensionalcontinuousfunction，MSCTSalgorithm，after parametersoptimization，canobtainnearly5timeshigherconvergencespeedandraises10ordersofmagnitudeontheaccu— racy．Therefore，MSCTSalgorithm ismoresuitableofrsolvingmultidimensionalcontinuousfunctionoptimization． Keywords：Tabusearchalgorithm；parameteroptimization；nei曲borh0