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:ItisprovedinthispaperthatcheckingatimedautomatonMwithrespecttoalineardurationpropertyDcanbedonebyinvestigatingonlytheintegraltimedstatesofM,Anequivalencerelationisintroducedinthispapertodividetheinfinitenumberofintegraltimedstatesintofinitenumberofequivalenceclasses.Basedonthis,amethodisproposedforcheckingwhetherMsatisfiesD.Insomecases,thenumberofequivalenceclassesistoolargeforacomputertomainpulate,Atechniqueforreducingthesearch-spaceforcheckinglineardurationpropoertyisalsodescribed.Thistechniqueismoresuitableforthecaseinthispaperthanthoseintheliteraturebecausemostofthosetechniquesaredesignedforreachablilityanalysis.
简介:Currently,schemaintegrationframeworksuseapproacheslikerule-based,machinelearning,etc.Thispaperpresentsanontology-basedwrapper-mediatorframeworkthatusesboththerule-basedandmachinelearningstrategiesatthesametime.Theproposedframeworkusesglobalandlocalontologiesforresolvingsyntacticandsemanticheterogeneity,andXMLforinteroperability.Theconceptsinthecandidateschemasaremergedonthebasisofthesimilaritycoefficient,whichiscalculatedusingthedefinedrulesandthepriormappingsstoredinthecase-base.
简介:计算机图形和人的计算机相互作用的许多最近的应用程序作为输入设备采用了颜色照相机和深度照相机。因此,拿不同颜色和深度的硬件的两种类型的有效刻度输入被要求。我们的途径移开在明确地象在深度和颜色照相机之间的转变一样解决照相机intrinsics的以前的方法使用非线性的优化的数字困难。混合参数的一个矩阵被介绍线性化我们的优化。混合参数从深度提供转变参量的空格(深度照相机图象)到一种颜色由从深度照相机联合深度照相机和旋转转变的内在的参数到颜色的参量的空格(颜色照相机图象)照相机。旋转转变和内在的参数能明确地在标准QRfactorisation的帮助下从我们的混合参数被计算。我们与地面真相深度信息被微软Kinect捕获的综合数据和真实世界的数据测试我们的算法。当由于使用混合参数的优点花少得多计算时间(Herreras方法的1/50和Raposos方法的1/10)时,实验证明我们的途径能向刻度的可比较的精确性提供最先进的算法。
简介:在2013年6月,美国国家安全机构分别地建议了小块零,叫的SIMON和斑点的二个家庭。这些零被设计在硬件和软件平台上最优地表现。在这份报纸,我们主要在SIMON的各种各样的版本上介绍零关联的线性密码翻译法。由使用missin-the-middle途径,第一,我们构造零关联的线性distinguishersSIMON,和零关联的线性攻击基于关键恢复的小心的分析被介绍阶段。第二,多维的零关联的线性攻击被用来减少数据复杂性。我们的零关联的线性攻击比不可能的微分攻击求婚由的更好表现卧病在床等。在ePrint报告2013/568。最后,我们也使用divide-and-conquer技术改进Javad等建议的线性密码翻译法的结果。在ePrint报告2013/663。
简介:Inthispaper,wepresentanovelapproachtosynthesizingfrontalandsemi-frontalcartoon-likefacialcaricaturesfromanimage.Thecaricatureisgeneratedbywarpingtheinputfacefromtheoriginalfeaturepointstothecorrespondingexaggeratedfeaturepoints.A3Dmeanfacemodelisincorporatedtofacilitatefacetocaricaturesbyinferringthedepthof3Dfeaturepointsandthespatialtransformation.Thenthe3Dfaceisdeformedbyusingnon-negativematrixfactorizationandprojectedbacktoimageplaneforfuturewarping.Toefficientlysolvethenonlinearspatialtransformation,weproposeanovelinitializationschemetosetupLevenberg-Marquardtoptimization.Accordingtothespatialtransformation,exaggerationisappliedtothemostsalientfeaturesbyexaggeratingtheirnormalizeddifferencefromthemean.Non-photorealisticrendering(NPR)basedstylizationcompletesthecartooncaricature.Experimentsdemonstratethatourmethodoutperformsexistingmethodsintermsofviewanglesandaestheticvisualquality.
简介:一线性(q,,,m(n))局部地可译码的代码(LDC)C:\mathbbF{\mathbbF}n\mathbbF{\mathbbF}m(n)是从向量空间\mathbbF的线性转变{\mathbbF}n到空间\mathbbF{\mathbbF}每消息标志xi能与概率为被恢复的m(n)至少\frac1|\mathbbF|+e\frac{1}{{\left|\mathbb{F}\right|}}从由查询仅仅q的一个使随机化的算法的C(x)的+\varepsilonC放(x),就算直到C的m(n)位置(x)被贿赂。在Dvir的一个最近的工作,为线性LDC降低界限的作者表演能为算术电路暗示更低的界限。他建议那证明界限更低因为在建筑群或真实的地上的LDC是为接近他的之一的一个好起点推测。我们的主要结果是m(n)=(n2)为在任何东西上的线性3质问LDC的更低的界限,可能无限,地。常数在(