简介:Inthispaper,weproposeadimension-reducing,K-meanclusteringprocedurebyProjectionPursuit(PP)techniquesoastoexploretheclusteringstructureofdatainhigh-dimensionalspaceintermsoflow-dimensionalprojectivepointsofdata,andweobtainthea.s.consistenceoftheestimatesoftheclustercentersandprojectionorientations.
简介:TheDeGrootmodelisoneofthemostclassicalmodelsinthefieldofopiniondynamics.ThestandardDeGrootmodelassumesthatagentsarehomogeneousandupdatetheiropinionsinthedirectionofaweightedaverageoftheirneighbors'opinions.Onenaturalquestioniswhetherasecondtypeofagentscouldsignificantlychangethemainpropertiesofthemodel.Theauthorsaddressthisquestionbyintroducingrebels,whoupdatetheiropinionstowardtheoppositeoftheirneighbors'weightedaverage.Theauthorsfindthattheexistenceofrebelsremarkablyaffectstheopiniondynamics.Undercertainmildconditions,theexistenceofafewrebelswillleadthegroupopiniontothegoldenmean,regardlessoftheinitialopinionsoftheagentsandthestructureofthelearningnetwork.ThisresultiscompletelydifferentfromthatofthestandardDeGrootmodel,wherethefinalconsensusopinionisdeterminedbyboththeinitialopinionsandthelearningtopology.Thestudythenprovidesnewinsightsintounderstandinghowheterogeneousindividualsinagroupreachconsensusandwhythegoldenmeanissocommoninhumansociety.
简介:Thispaperstudiesadynamicmean-varianceportfolioselectionproblemwithrandomliabilityintheaffineinterestrateenvironment,wherethefinancialmarketconsistsofthreeassets:onerisk-freeasset,oneriskyassetandonezero-couponbond.AssumethatshortrateisdrivenbyaffineinterestratemodelandliabilityprocessisdescribedbythedriftedBrownianmotion,inaddition,stockpricedynamicsisaffectedbyinterestratedynamics.Theinvestorsexpecttolookforanoptimalstrategytominimizethevarianceoftheterminalsurplusforagivenexpectedterminalsurplus.TheefficientstrategyandtheefficientfrontierareexplicitlyobtainedbyapplyingdynamicprogrammingprincipleandLagrangedualitytheorem.Anumericalexampleisgiventoillustrateourresultsandsomeeconomicimplicationsareanalyzed.
简介:这份报纸讨论吝啬地的向后的随机的微分方程(吝啬地的BSDE)与跳并且控制吝啬地的BSDE与的一种新类型跳,也就是吝啬地的BSDE与强烈跳结合了联系控制问题的价值功能。作者首先为BSDE的上述二种类型证明存在和唯一以及一条比较定理。为这,作者使用一个近似方法。在Peng在1997介绍的随机的向后的semigroups的观点的帮助下,然后,作者为值函数得到动态编程原则(DPP)。而且,作者证明值函数是integro部分的微分方程,它在连续函数的一个足够的空格唯一由Barles介绍了的联系非局部的Hamilton-Jacobi-Bellman(HJB)的一个粘性解决方案,等。在1997。