简介:这份报纸学习nonconvex的一个班的系统稳定性问题微分包括。起初,基本稳定性结果被优点获得局部地,Lipschitz连续Lyapunov工作。而且,一个概括不变性原则和相关吸引力条件被建议并且证明了由于凸状的缺席克服技术困难。在技术分析,新奇珍视集合的衍生物被建议处理nonsmooth系统和nonsmoothLyapunov功能。另外,获得的结果与在有常规Lyapunov功能的凸的微分包括的情况中的存在的一致。最后,解说性的例子被给显示出方法的有效性。
简介:Duetodramaticallyincreasinginformationpublishedinsocialnetworks,privacyissueshavegivenrisetopublicconcerns.Althoughthepresenceofdifferentialprivacyprovidesprivacyprotectionwiththeoreticalfoundations,thetrade-offbetweenprivacyanddatautilitystilldemandsfurtherimprovement.However,mostexistingstudiesdonotconsiderthequantitativeimpactoftheadversarywhenmeasuringdatautility.Inthispaper,wefirstlyproposeapersonalizeddifferentialprivacymethodbasedonsocialdistance.Then,weanalyzethemaximumdatautilitywhenusersandadversariesareblindtothestrategysetsofeachother.Weformalizeallthepayofffunctionsinthedifferentialprivacysense,whichisfollowedbytheestablishmentofastaticBayesiangame.Thetrade-offiscalculatedbyderivingtheBayesianNashequilibriumwithamodifiedreinforcementlearningalgorithm.Theproposedmethodachievesfastconvergencebyreducingthecardinalityfromnto2.Inaddition,thein-placetrade-offcanmaximizetheuser'sdatautilityiftheactionsetsoftheuserandtheadversaryarepublicwhilethestrategysetsareunrevealed.Ourextensiveexperimentsonthereal-worlddatasetprovetheproposedmodeliseffectiveandfeasible.
简介:Thispaperwillpresentanapproximate/adaptivedynamicprogramming(ADP)algorithm,thatusestheideaofintegralreinforcementlearning(IRL),todetermineonlinetheNashequilibriumsolutionforthetwo-playerzerosumdifferentialgamewithlineardynamicsandinfinitehorizonquadraticcost.Thealgorithmisbuiltaroundaniterativemethodthathasbeendevelopedinthecontrolengineeringcommunityforsolvingthecontinuous-timegamealgebraicRiccatiequation(CT-GARE),whichunderliesthegameproblem.W...
简介:原来的微分进化(DE)的控制参数被使全部进化进程全部固定了。然而,它不是到在为不同优化问题的DE的适当地设置的控制参数的一项容易的任务。根据选择在寻找的地方产生差别向量的二不同单个向量的相对位置,为差别向量的规模因素F的自我改编的策略被建议。以在当前的人口的目标向量的集中地位,自我改编的转线路概率常数CR策略被建议。因此,当更坏的目标向量有大CR时,好目标向量有更低的CR。同时,变化操作员被修改改进集中速度。这些建议途径的表演与一些基准问题的使用被学习;适用于一个三关节的冗余的操纵者计划的轨道。最后,实验结果证明建议途径能极大地改进坚韧性;集中速度。
简介:Inthispaper,weconstructacompositeMilsteinmethodfornonlinearstochasticdifferentialdelayequations.ThenweanalyzethemeansquarestabilityforthismethodandobtainthestepsizeconditionunderwhichthecompositeMilsteinmethodismeansquarestable.Moreover,wegetthestepsizeconditionunderwhichthecompositeMilsteinmethodisglobalmeansquarestable.Anonlinearteststochasticdifferentialdelayequationisgivenfornumericaltests.Theresultsofnumericaltestsverifythetheoreticalresultsproposed.
简介:许多真实世界的问题是动态的,要求能连续地追踪随着时间的过去改变optima(最佳)的优化算法。这份报纸用近邻居的效果的观点建议一个改进微分进化算法决定一位个人邻居,为追踪在动态环境的多重optima。用近邻居的效果的新变化策略也被介绍。它由在它的邻居利用存储记忆点,并且利用生产由的微分向量创造个人nearneighbor优异并且near-neighbor-inferior。从生物免疫系统拿灵感,一个免疫系统基于计划为很快检测并且对环境变化作出回应被介绍。另外,一个差别相关的multidirectional扩大计划被介绍从不同尺寸集成珍贵信息为有效地并且很快在搜索空间发现有希望的最佳。动态情形的实验由典型动态测试instancemoving山峰问题创造了,证明了近邻居并且基于的免疫系统微分进化算法(NIDE)在处理动态优化功能是有效的。
简介:Manypracticalsystemsinphysicalandtechnicalscienceshaveimpulsivedynamicalbehaviorsduringtheevolutionprocesswhichcanbemodeledbyimpulsivedifferentialequations.Inthispaper,weprovetheapproximatecontrollabilityofcontrolsystemsgovernedbyaclassofimpulsiveneutralstochasticfunctionaldifferentialsystemwithstate-dependentdelayinHilbertspaces.Sufficientconditionsforapproximatecontrollabilityofthecontrolsystemsareestablishedunderthenaturalassumptionthatthecorrespondinglinearsystemisapproximatelycontrollable.Theresultsareobtainedbyusingsemigrouptheory,stochasticanalysistechniques,fixedpointapproachandabstractphasespaceaxioms.Anexampleisprovidedtoillustratetheapplicationoftheobtainedresults.