简介:广告(广告)选择在赞助搜索起一个重要作用,自从它是一个在上游的部件并且将重重地影响随后的拍卖机制的有效性。然而,大多数存在广告选择方法认为广告选择是一个相对独立的模块,并且仅仅考虑文字或在广告选择过程期间的在询问和关键词之间的语义匹配。在这份报纸,我们主张这条途径不是全球性最佳的。我们的建议是作为如此的一个优化问题提出广告选择,选择广告能和下游的部件工作(例如,拍卖机制)完成用户按的最大化,广告客户聚会福利,并且搜索引擎收入(我们作为市场目的叫这些客观函数的联合便于引用)。到这个目的,我们1)提取一捆特征代表每询问和关键词,并且2)训练印射特征到显示关键词是否被选择的一个二进制变量的一个机器学习模型,由最大化上述的市场目的。这形式化似乎相当自然;然而,因为市场目的是非凸的,它是技术上困难的,不连续,并且由于评价和秒价格考虑模型参数的indifferentiable在拍卖机制统治。处理挑战,我们建议市场目的概率的近似,它是光滑的并且能被常规优化技术有效地优化。我们测试从一个商业搜索引擎用赞助搜索日志与我们的建议方法学习的广告选择模型。试验性的结果证明我们的方法罐头显著地在调查下面在所有度量标准上超过几个广告选择算法。
简介:Thispaperreportsawind-poweredwaterpumpingsystemimplementedinruralsideofPakistan.ThedesignmethodologypresentedinthispapershallenhancetheefficiencyoftheexistingSavoniusturbine’sperformancebymodifyingitaerodynamically.BladeswithdifferenttwistanglesaredesignedinSolidEdgeandanalyzedusingcomputationalfluiddynamics(CFD)withANSYSFLUENTsoftware.Staticandrotationalanalysesareperformedtogetoptimizedtwistangleandresultsarehighlighted.Theperformancesoftheturbineinbothstaticandrotationalanalysesarecompared.
简介:TheVacuumVessel(VV)systemisanessentialcomponentofKedaTorusforexperiment(KTX),andvariousscenariosmighttakeplaceonit.TheVV’ssupportsshouldbeadequatelystrongtostandagainstvariousloadsonVV,whichmighthappeninextremescenarios.Therefore,thedesignofVVsupportsisverifiedinasingleextremescenarioandissubsequentlyoptimizedinthisreport.ThenumericalsimulationbasedonFiniteElementtheoryisperformedasthemajormethodforanalysisandoptimization.Theelectromagneticforceinpreviousanalysesservesastheloadforthemechanicalanalysesofsupports.Duringtheoptimization,thestressesoftheWsupportsdecreaseremarkablyafterintroducingcotters.Finally,theoptimumdesignhasbeenworkedout.Itsatisfiestherequirementsregardingthestrengthandconvenienceinassembly.
简介:Theresearchacceleratorsaregrowinginenergyandpowerwhichinducesanincreaseoftheprojectcosts(Fig.1).TheRFcavitiesandpowersupplymainlycontributethecostofthesuperconductinglinac,whichisacompetitivetechnologyforhighpowermachine.AcodewithoptimizationalgorithmisdevelopedtooptimizethebetavalueofSCcavityfamilyandtransitionenergytoincreasetheTTFvalueofthewholelinac.
简介:Duringtheingotcastingprocess,theparametricchangesofthemoltensteelpouringprocessoftencausebothingotsurfaceandinternalqualitydefects.Withtheincreasinglywideapplicationofcomputersinindustrialtechnology,numericalsimulationsofthemoltensteelpouringprocessenablepredictionsofthedefectsthatwillariseduringthepouringandsolidificationprocess,especiallyforthemoltensteelflowduringtheearlystagesofcasting.Inaddition,usingacomputertosimulatethecastingprocessofa40tingotandsolidificationcanpreventmanydefects.Thesimulationresultsindicatethatcontrollingthepouringspeedtolessthan80kg/s,hangingthecastingpowderinamoldwiththehangingheightmaintainedat100-150mm,andcontrollingthetaperangleofthemoldbrickoutlettoasuitablelevelcanreducetheprobabilityofslagentrapmentandimprovetheinternalandsurfacequalityoftheingot.
简介:无线网络在更宽的光谱利用的时尚下面被开发(例如,认知无线电)并且多跳跃通讯(例如,无线网孔网络)。在这些范例,怎么有效地与最小化的相互的干扰分配光谱到不同传播连接成为关键担心。在这份报纸,我们在认知收音机网络(CRN)经由光谱分配学习产量优化。以前的研究合并冲突图或SINR模型描绘干扰关系。然而,以前的模型忽视积聚的干扰效果并且导致讨厌的干扰和非最优的结果,当工作在所有潜在的连接之中在估计的RSS(收到的信号力量)的精确性上基于后者模型忽视它的重信赖时。两个是不适当的描绘在干扰和产量之间的复杂关系。到这个目的,由考虑CR的特征,象光谱差异和间断OFDM一样,我们建议一个帮助测量的基于SINR的跨层的产量优化答案。我们的工作在不同的层使特征担心:在物理层,我们在场改进SINR模型的精确性的一个有效RSS评价算法;在上面的层,流动水平为WMN的基于SINR的产量优化问题作为一个混合整数被建模被证明NP难的非线性的编程问题。解决这个问题,一集中(1)最佳的算法和一个有效分布式的算法被提供。评估算法表演,真实世界的踪迹被用来说明我们的计划的有效性。
简介:Theshiphullsurfaceoptimizationbasedonthewaveresistanceisanimportantissueintheshipengineeringindustry.Thewaveletmethodmayprovideaconvenienttoolforthesurfacehulloptimization.Asapreliminarystudy,weusethewaveletmethodtooptimizethehullsurfacebasedontheMichelwaveresistanceforaWigleymodelinthispaper.Firstly,weexpressthemodel’ssurfacebythewaveletdecompositionexpressionsandobtainareconstructedsurfaceandthenvalidateitsaccuracy.Secondly,werewritetheMichelwaveresistanceformulainthewaveletbases,resultinginasimpleformulacontainingonlytheshiphullsurface’swaveletcoefficients.Thirdly,wetakethesewaveletcoefficientsasoptimizationvariables,andanalyzethemainwaveresistancedistributionintermsofscalesandlocations,toreducethenumberofoptimizationvariables.Finally,weobtaintheoptimalhullsurfaceoftheWigleymodelthroughgeneticalgorithms,reducingthewaveresistancealmostbyahalf.Itisshownthatthewaveletmethodmayprovideanewapproachforthehulloptimization.
简介:Flocculationtimeisconventionallybelievedtobeproportionaltotheflocculationefficiencyofacylindricalfluidizedbedflocculator.However,inasingle-stagevelocitygradientsituation,theflocculationefficiencydecreaseswhentheoptimalflocculationtimeisexceeded.Amulti-stagevelocitygradientwasestablishedinacylindricalfluidizedbedflocculator,basedonthehydraulicclassificationtheory.Thismulti-stagevelocitygradientfluidizedbedflocculator(MGF)createdamoresuitableenvironmentforflocgrowthandprotection,whichwasconfirmedbythesizedistributionofflocsalongthebedheight.Correspondingly,theabatementefficienciesforKaolinslurryanddyedwastewatertreatmentintheMGFwereenhancedby5-10%,andby7-20%,respectively,comparedwiththoseinthesingle-stagevelocitygradientfluidizedbedflocculators(SGFs).TheinitialbedheightdistributionratioalongthevelocitygradientswasanimportantfactorforMGFoptimization.
简介:城市的总线网络的多客观的优化能帮助改进运输系统的操作效率并且为在中国减少城市的交通拥挤开发策略。工作使用了累积前景理论,当前为在无常下面的决定的最有影响的模型,优化城市的公共汽车网络。完成研究目的,工作开发了城市的总线网络优化,包括的优化原则,优化目的和限制的理论框架。而且,优化目的能包括地从时间,空间和价值的尺寸反映旅客和总线公司的期望。它在以前的研究独自与仅仅一个股东或尺寸相比更科学、合理。另外,为由到理想的答案TOPSIS的类似的顺序偏爱的技术被用来决定积极、否定的理想的选择。在优化选择和理想的选择之间的关联被灰关系分析同时估计。累积前景理论CPT被用来由比较每种选择的全面前景价值决定最好的选择,精确地在实际生活与期望的用途理论相比描述决策行为。最后,Xian的大小写证明方法能更好调整总线网络,并且优化解决方案是更合理的满足实际需要。
简介:Thealternatingdirectionmethodofmultipliers(ADMMforshort)isefficientforlinearlyconstrainedconvexoptimizationproblem.ThepracticalcomputationalcostofADMMdependsonthesub-problemsolvers.Theproximalpointalgorithmisacommonsub-problem-solver.However,theproximalparameterissensitiveintheproximalADMM.Inthispaper,weproposeahomotopy-basedproximallinearizedADMM,inwhichahomotopymethodisusedtosolvethesub-problemsateachiteration.Undersomesuitableconditions,theglobalconvergenceandtheconvergencerateofO(1/k)intheworstcaseoftheproposedmethodareproven.Somepreliminarynumericalresultsindicatethevalidityoftheproposedmethod.
简介:Itisimportanttounderstandtheimpactofwing-morphingonaerodynamicperformanceinthestudyofflapping-wingflightofbirdsandinsects.Weuseaflappingplatehingedwithatrailing-edgeflapasasimplifiedmodelforflexible/morphingwingsinhovering.Thetrailing-edgeflappingmotionisoptimizedbyanadjoint-basedapproach.Theoptimizedconfigurationsuggeststhatthetrailing-edgeflapcansubstantiallyenhancetheoveralllift.Furtheranalysisindicatesthattheliftenhancementbythetrailingedgeflappingisfromthechangeofcirculationintwoways:thelocalcirculationchangebytherotationalmotionoftheflap,andthemodificationofvortexsheddingprocessbytherelativelocationbetweenthetrailing-edgeflapandleading-edgemainplate.
简介:Performance-baseddesigninearthquakeengineeringisastructuraloptimizationproblemthathas,astheobjective,thedeterminationofdesignparametersfortheminimizationoftotalcosts,whileatthesametimesatisfyingminimumreliabilitylevelsforthespecifiedperformancecriteria.Totalcostsincludethoseforconstructionandstructuraldamagerepairs,thoseassociatedwithnon-structuralcomponentsandthesocialcostsofeconomiclosses,injuriesandfatalities.Thispaperpresentsageneralframeworktoapproachthisproblem,usinganumericaloptimizationstrategyandincorporatingtheuseofneuralnetworksfortheevaluationofdynamicresponsesandthereliabilitylevelsachievedforagivensetofdesignparameters.Thestrategyisappliedtoanexampleofathree-storyofficebuilding.Theresultsshowtheimportanceofconsideringthesocialcosts,andtheoptimumfailureprobabilitieswhenminimumreliabilityconstraintsarenottakenintoaccount.
简介:Amethodforoptimizingthespectraldistortionofanultrafastpulseinapolarization-maintainingpicosecondlinear-cavityfiberlaserwithaone-stagefiberamplifierisproposedanddemonstrated.Themechanismofcontrolofthespectraldistortioninthefibersystemhasbeeninvestigated.TheexperimentalandtheoreticalresultsillustratethatthefilteringeffectofafiberBragggratingcaneffectivelydecreasethespectraloscillatorydistortionaccumulatedbyself-phasemodulation.InjectedintoaNd:YAGregenerativeamplifier,theultrafastpulsecouldproducehighpulseenergyof1.2mJatarepetitionrateof1kHz.