简介：Asemi-structureddocumenthasmorestructuredinformationcomparedtoanordinarydocument,andtherelationamongsemi-structureddocumentscanbefullyutilized.Inordertotakeadvantageofthestructureandlinkinformationinasemi-structureddocumentforbettermining,astructuredlinkvectormodel(SLVM)ispresentedinthispaper,whereavectorrepresentsadocument,andvectors'elementsaredeterminedbyterms,documentstructureandneighboringdocuments.TextminingbasedonSLVMisdescribedintheprocedureofK-meansforbriefnessandclarity:calculatingdocumentsimilarityandcalculatingclustercenter.TheclusteringbasedonSLVMperformssignificantlybetterthanthatbasedonaconventionalvectorspacemodelintheexperiments,anditsFvalueincreasesfrom0.65-0.73to0.82-0.86.

简介：Currently,schemaintegrationframeworksuseapproacheslikerule-based,machinelearning,etc.Thispaperpresentsanontology-basedwrapper-mediatorframeworkthatusesboththerule-basedandmachinelearningstrategiesatthesametime.Theproposedframeworkusesglobalandlocalontologiesforresolvingsyntacticandsemanticheterogeneity,andXMLforinteroperability.Theconceptsinthecandidateschemasaremergedonthebasisofthesimilaritycoefficient,whichiscalculatedusingthedefinedrulesandthepriormappingsstoredinthecase-base.

简介：Theadaptivecriticheuristichasbeenapopularalgorithminreinforcementlearning(RL)andapproximatedynamicprogramming(ADP)alike.ItisoneofthefirstRLandADPalgorithms.RLandADPalgorithmsareparticularlyusefulforsolvingMarkovdecisionprocesses(MDPs)thatsufferfromthecursesofdimensionalityandmodeling.Manyreal-worldproblems,however,tendtobesemi-Markovdecisionprocesses(SMDPs)inwhichthetimespentineachtransitionoftheunderlyingMarkovchainsisitselfarandomvariab...

简介：数据安全为云计算是领先的担心和主要挑战之一。这个问题与云计算的发展正在变得越来越严肃。然而，存在保存隐私的数据分享技术也没能阻止隐私的漏或招致信息损失的巨大的数量。在这份报纸，我们建议一种新奇技术，作为基于连接的匿名模型称为，它与有一种尺寸的伪标识符组(QI组)一起完成K匿名不到K。同时，半同质的归纳被介绍对同质的归纳招致的攻击。为了实现基于连接的anonymization，当模特儿，我们建议一个简单还有效的启发式的本地人重新代码方法。真实数据集的广泛的实验也被进行证明实用程序被我们的途径显著地与最先进的方法相比改进了。

简介：Inthispaper,wepresentanovelapproachtosynthesizingfrontalandsemi-frontalcartoon-likefacialcaricaturesfromanimage.Thecaricatureisgeneratedbywarpingtheinputfacefromtheoriginalfeaturepointstothecorrespondingexaggeratedfeaturepoints.A3Dmeanfacemodelisincorporatedtofacilitatefacetocaricaturesbyinferringthedepthof3Dfeaturepointsandthespatialtransformation.Thenthe3Dfaceisdeformedbyusingnon-negativematrixfactorizationandprojectedbacktoimageplaneforfuturewarping.Toefficientlysolvethenonlinearspatialtransformation,weproposeanovelinitializationschemetosetupLevenberg-Marquardtoptimization.Accordingtothespatialtransformation,exaggerationisappliedtothemostsalientfeaturesbyexaggeratingtheirnormalizeddifferencefromthemean.Non-photorealisticrendering(NPR)basedstylizationcompletesthecartooncaricature.Experimentsdemonstratethatourmethodoutperformsexistingmethodsintermsofviewanglesandaestheticvisualquality.

简介：Semi-TCP协议将传统TCP的拥塞控制下放到MAC层,构成逐跳的拥塞控制来判断节点间的拥塞状态,它比传统TCP以及在这基础上改进的无线TCP更适应于动态拓扑的网络。但是目前对Semi-TCP协议的研究还仅局限于计算机仿真阶段。因此尝试基于实际平台RP4412开发板,不修改已有MAC层的控制帧和已有的硬件配置信息,利用软件定义发送数据帧的形式,在帧体域携带节点缓存状态量化信息,完成节点拥塞状态信息的传递,以此实现并测试Semi-TCP在实际平台下拥塞控制的性能。