简介:Thebilevelprogrammingisappliedtosolvehierarchicalintelligencecontrolproblemsinsuchfieldsasindustry,agriculture,transportation,military,andsoon.Thispaperpresentsaquadraticobjectivepenaltyfunctionwithtwopenaltyparametersforinequalityconstrainedbilevelprogramming.Undersomeconditions,theoptimalsolutiontothebilevelprogrammingdefinedbythequadraticobjectivepenaltyfunctionisprovedtobeanoptimalsolutiontotheoriginalbilevelprogramming.Moreover,basedonthequadraticobjectivepenaltyfunction,analgorithmisdevelopedtofindanoptimalsolutiontotheoriginalbilevelprogramming,anditsconvergenceprovedundersomeconditions.Furthermore,undertheassumptionofconvexityatlowerlevelproblems,aquadraticobjectivepenaltyfunctionwithoutlowerlevelproblemsisdefinedandisprovedequaltotheoriginalbilevelprogramming.
简介:Inthispaper,arecursivequadraticprogrammingalgorithmisproposedandstudied.ThelinesearchfunctionsusedaxeHan'snondifferentiablepenaltyfunctionswithasecondoederpenaltyterm.InordertoavoidmaratoseffectFukushima'smixeddirectionisusedasthedirectionoflinesearch.Finally,weprovetheglobalconvergenceandthelocalsecondorderconvergenceofthealgorithm.
简介:在这份报纸,我们调查为剩余进程的打折的惩罚功能由piecewise描述了的Gerber-Shiu确定的Markov进程(PDMP).We导出一个不可分的方程因为Gerber-Shiu打折惩罚功能,并且当起始的剩余是零时,获得准确解决方案。Dickson公式也被概括到现在的剩余过程。关键词Gerber-Shiu打折惩罚功能-piecewise确定的Markov过程-最终的毁灭概率-Volterra不可分的方程先生(2000)题目分类91B30-62P20由中国的国家自然科学基础支持了(资助Nos.10926161,10901086,10871102),2007CB814905和研究为高等教育的博士程序资助的中国(973程序)的国家基本研究节目
简介:Thispaperaimstodevelopapowerpenaltymethodforalinearparabolicvariationalinequality(Ⅵ)intwospatialdimensionsgoverningthetwo-assetAmericanoptionvaluation.Thismethodyieldsatwo-dimensionalnonlinearparabolicPDEcontainingapowerpenaltytermwithpenaltyconstantλ>1andapowerparameterk>0.WeshowthatthenonlinearPDEisuniquelysolvableandthesolutionofthePDEconvergestothatoftheⅥattherateoforder(?)(λ~(-k/2)).AfittedfinitevolumemethodisdesignedtosolvethenonlinearPDE,andsomenumericalexperimentsareperformedtoillustratetheusefulnessofthismethod.
简介:AbstractInthisarticlewebrieflystudiedthefictitiousdomainmethodsforsteadyandnonsteadyNavier-StokesequationsbasedonPenaltyfactorεontheextendeddomain.Theconvergenceu?→uinH01(Ω)dandL2(Ω)d(d=2,3)isgivenaswellasp?→pinL02(Ω).
简介:Inordertoimprovetheperformanceofsupportvectormachine(SVM)applicationsinthefieldofcreditriskevaluation,anadaptiveLqSVMmodelwithGausskernel(ALqG-SVM)isproposedtoevaluatecreditrisks.Thenon-adaptivepenaltyoftheobjectfunctionisextendedto(0,2]toincreaseclassificationaccuracy.Tofurtherimprovethegeneralizationperformanceoftheproposedmodel,theGausskernelisintroduced,thusthenon-linearclassificationproblemcanbelinearlyseparatedinhigherdimensio...
简介:InthispaperwestudythesolutionofSATproblemsformulatedasdiscretedecisionanddiscreteconstrainedoptimizationproblems.Constrainedformulationsarebetterthantraditionalunconstrainedformulationsbecauseviolatedconstraintsmayprovideadditionalforcestoleadasearchtowardsasatisfiableassignment.Wesummarizethetheoryofextendedsaddlepointsinpenaltyformulationsforsolvingdiscreteconstrainedoptimizationproblemsandtheassociateddiscretepenaltymethod(DPM).Wethenexaminevariousformulationsoftheobjectivefunction,choicesofneighborhoodinDPM,strategiesforupdatingpenalties,andheuristicsforavoidingtraps.ExperimentalevaluationsonhardbenchmarkinstancespinpointthattrapscontributesignificantlytotheinefficiencyofDPMandforceatrajectorytorepeatedlyvisitthesamesetofornearbypointsintheoriginalvariablespace.Toaddressthisissue,weproposeandstudytwotrap-avoidancestrategies.Thefirststrategyaddsextrapenaltiesonunsatisfiedclausesinsideatrap,leadingtoverylargepenaltiesforunsatisfiedclausesthataretrappedmoreoftenandmakingtheseclausesmorelikelytobesatisfiedinthefuture.Thesecondstrategystoresinformationonpointsvisitedbefore,whetherinsidetrapsornot,andavoidsvisitingpointsthatareclosetopointsvisitedbefore.Itcanbeimplementedbymodifyingthepenaltyfunctioninsuchawaythat,ifatrajectorygetsclosetopointsvisitedbefore,anextrapenaltywilltakeeffectandforcethetrajectorytoanewregion.Itspecializestothefirststrategybecausetrapsarespecialcasesofpointsvisitedbefore.Finally,weshowexperimentalresultsonevaluatingbenchmarksintheDIMACSandSATLIBarchivesandcompareourresultswithexistingresultsonGSAT,WalkSAT,LSDL,andGrasp.TheresultsdemonstratethatDPMwithtrapavoidanceisrobustaswellaseffectiveforsolvinghardSATproblems.
简介:
简介:Ⅰ.RestrictionandAnalysisofDeathPenaltyinExistingSubstantiveLawinChinaInthenormativecriminallaw,thepersonaldangerousnessisthepossibilityoftheactortocommitacrimeagain.Beforethepersonaldangerousnessastheunaccomplishedcrimedoesnotbecomeaconsensusconcepttoentertheviewofthepublicscholars,howdoesitfitwiththecontrolofdeathpenaltyinourexistingsubstantivelaw?
简介:Inthispaperwedescribeamulti-gridalgorithmformixedproblemswithpenaltybythelinearfiniteelementapproximation.Itisprovedthattheconvergencerateofthealgorithmisboundedawayfrom1independentlyofthemeshsize.Forconvenience,weonlydiscussJacobirelaxationassmoothingoperatorindetail.
简介:Forill-posedbilevelprogrammingproblem,theoptimisticsolutionisalwaysthebestdecisionfortheupperlevelbutitisnotalwaysthebestchoiceforbothlevelsiftheauthorsconsiderthemodel'ssatisfactorydegreeinapplication.Toacquireamoresatisfyingsolutionthantheoptimisticonetorealizethetwolevels'mostprofits,thispaperconsidersbothlevels'satisfactorydegreeandconstructsaminimizationproblemofthetwoobjectivefunctionsbyweightedsummation.Then,usingthedualitygapofthelowerlevelasthepenaltyfunction,theauthorstransferthesetwolevelsproblemtoasingleoneandproposeacorrespondingalgorithm.Finally,theauthorsgiveanexampletoshowamoresatisfyingsolutionthantheoptimisticsolutioncanbeachievedbythisalgorithm.
简介:我们建议错误与异构、各向异性的散开为加权的内部惩罚估计advection-diffusion-reaction方程的不连续的Galerkin近似并且分析后验精力标准。重量,在分析起一个关键作用,取决于散开张肌并且被用来在不连续的Galerkin方法提出一致性术语。错误上面的界限,所有常数在被指定,由三个术语组成:仅仅取决于元素的一个剩余评估者分离溶液剩余,在方法使用的重量明确地进入的一个散开的流动评估者,和是因为不连续的有限元素空格的使用的非零的一个非遵守的评估者的明智的变化。三个评估者能被近似错误局部地围住。注意在本地人在常数的问题参数上被给相关性的详细规格降低错误界限。为中等移流,关于散开异质的完整的坚韧性在不连续的Galerkin方法由于重量的特定的设计被完成,这被显示出,当时完全本地的散开anisotropies遗体和影响通过条件的方形的根的常数散开张肌数。为主导的移流,它被显示出,在由连续有限元素上的Verfürth的以前的工作的精神,本地人降低错误界限,这能被写,常数包含一为本地网孔尺寸到的比率截止散开张肌的最低本地eignevalue的方形的根相互。
简介:Atunableslowlightof2.5-Gb/spseudo-randombinarysequencesignalusinga1550-nmvertical-cavitysurface-emittinglaser(VCSEL)isexperimentallydemonstrated.Theinfluencesofthebiascurrentandthegainsaturationontheslowlightareinvestigated.Withbiascurrentincreasing,tunableopticalgroupdelayupto98psisobtainedatroomtemperature.Demonstrationofthetimedelaybetween16and24psbysignalintensitychangeisreported.Underanappropriatebiascurrent,bytuningtheinputsignaltotrackthepeakgainwavelengthoftheVCSEL,slowlightofapowerpenaltyaslowas1dBisachieved.Withsuchalowpowerpenalty,theVCSELhasagreatpotentialapplicationasacompactopticalbuffer.
简介:这份报纸涉及为一个维的充分非线性的秒顺序开发精确、有效的数字方法椭圆形、寓言的部分微分方程(PDE)。在纸我们在场为构造高顺序内部惩罚的一个一般框架为这些充分非线性的PDE的接近的粘性答案的不连续的Galerkin(IP-DG)方法。为了捕获解决方案的第二顺序衍生物uxx的断绝,u,三独立函数p1,p2和p3被介绍用各种各样的片面限制代表数字衍生物。建议DG框架,基于内在的PDE的非标准的混合明确的表达,把一个非线性的问题嵌进非线性被修改了包括第二顺序衍生物uxx的多重价值的方程的一个主要线性的系统。建议框架扩大有限差别框架由作者开发了在的一个同伴[9]并且允许用高顺序多项式和不一致的网孔的充分非线性的PDE的近似。除了非标准的混合明确的表达设置,另一个主要想法是由与微分操作符一致并且满足某些monotonicity(叫的g-monotonicity)的一个数字操作符代替充分非线性的微分操作符性质。保证如此的g-monotonicity,构造的关键是介绍数字时刻,它在建议DG框架起一个关键作用。g-monotonicity给DG方法能力选择算术地“改正”答案(即,粘性答案)在所有可能的答案之中。而且,g-monotonicity允许更有效的非线性的解答者的可能的发展能作为代数学的系统的特殊非线性被探索到decouple方程。这份报纸也为习惯于guage的几个数字测试问题论述并且分析数字结果建议DG方法的精确性和效率。[从作者抽象]