简介:Fortheimageofasmoothsurfaceobjectfullycontainedwithinthefieldofviewandilluminatedinandarbitrarydirection,thispaperdiscussestheexistenceanduniquenessotheconditionsforsolvingashape-from-shadingproblemundertheconditionsthattheFourierseriesexpansionoftheimageintensitycontainsonlyzeroandfirstordertermsinapolarcoordinatesystem.Threetheoremsareestablished,onefortheexistenceandtwofortheuniquenessofz-axissymmetricshapefromshading.
简介:Inrecentyears,wehavebeendevelopinganewframeworkforacquiringmedicalknowledgefromEncyclopedictexts.Thisframeworkconsistsofthreemajorparts.Thefirstpartisanextendedhigh-levelconceptuallanguage(calledHLCL1.1)forusebyknowledgeengineerstoformalizeknowledgetextsinanencyclopedia.TheotherpartisanHLCL1.1compilerforparsingandanalyzingtheformalizedtextsintoknowledgemodels.Thethirdpartisasetofdomain-specificontologiesforsharingknowledge.
简介:采矿诱发性是必要的提供诊断。这研究瞄准提取在多重句子或EDU(基本讲话单位)以内存在的诱发性。因为他们以某个方式成为明确,研究强调诱发性动词的使用一个原因的作为结果的事件,例如,“蚜虫从米饭叶子吮吸傻瓜。然后,叶子将缩小。后来,他们将变得黄;干燥。'.一个动词能也是在原因之间的原因动词的连接;在EDU以内完成,例如,“蚜虫从引起叶子被缩小的米饭叶子吮吸傻瓜”(“引起”用泰语等价于一个原因动词的连接)。研究面对二个主要问题:从文件识别有趣的诱发性事件;识别他们的边界。然后,我们由使用二种不同机器学习技术在动词上建议采矿,中间广场Bayes;支持向量机。结果的采矿规则将被用于鉴定;从文本的多重EDU的诱发性抽取。我们的多重EDU抽取从中间广场Bayes与0.75召回显示出0.88精确;有从支持向量机的0.76召回的0.89精确。
简介:Thispaperpresentsafunctionallanguagefortheunambiguousdescriptionofdigitalcircuits,amethodandalgorithmstoobtainastandard-celllayout,andacomparativeevaluationofthedevelopedfunctionalstandard-cellplacementtechnique.Thepresentedplacementschemeisdifferentfromtraditionalmethodsbecausethecompletelayoutgrometryisspecifiedandconstructedautomaticallyfromafunctionaldescription.Theconstructionreliesonatranslationthatcombinesthesimplicityofstandard-cellswiththeeleganceoffunctionalprogramming.Anevaluationofthemethodintroducedshowsthatthequalityoftheresultingplacementisclosetotheresultsachievedwithsimulatedannealingwhilethecomputationtimeissignificantlyless.Furthermore,theevaluationsuggeststoemploythefunctionalplacementmethodinconjunctionwithlow-temperaturesimulatedannealingforrunning-timereductionandimprovedresults.
简介:这篇论文介绍一个测量过程,叫了RmFFP,它为建模描述一套操作并且用OO方法需求模型从高级说明估计面向对象的软件系统的尺寸。OO方法是一个自动软件生产方法。这个工作的贡献是系统地定义允许在软件生产过程的一个早阶段估计功能的尺寸的一套规则,根据COSMIC-FFP。做这,我们描述为软件测量跟随一个过程模型的步骤的建议测量过程的设计,应用程序,和分析。我们也以它的重制度在RmFFP的评估上报导起始的结果。电子增补材料这篇文章(doi:10.1007/s11390-007-9050-6)的联机版本contatins增补材料,它对授权用户可得到。
简介:ThispaperproposesanapproachtosynthesizefunctionalprogramsofBackus'FPsystem^[1,2]frominput/outputinstances.Basedonatheoryoforthogonalexpansionofprograms^3[,4],thetaskofprogramsynthesisisexpressedinprogramequations,andfulfilledbysolvingthemaccordingtotheknowledgeabouttheequivalencebetweenprograms.Somegeneralknowledgeofsolvingprogramequationswithanumberofexamplesaregiveninthepaper.
简介:RoutingprotocolsplayanimportantroleintheInternetandthetestrequirementsarerunningup.Totestroutingprotocolsmoreefficiently,severalenhancingtechniquesareappliedintheprotocolintegratedtestsystemdescribedinthispaper.TheImplementationUnderTestismodeledasablackboxwithwindows.Thetestsystemisendowedwithmultiplechannelsandmultipleportstotestdistributedprotocols.Thetestsuiteandotherrelatedaspectsarealsoextended.Meanwhile,thepassivetestingisintroducedtotest,analyzeandmanageroutingprotocolsintheproductionfield,whichisabletoperformtheconformancetest,theinteroperabilitytestandtheperformancetest.Thestatemachineofpeersessionsistestedwiththestatesynchronizationalgorithm,andtheroutinginformationmanipulationandotheroperationsarecheckedandanalyzedwiththemethodslikethetopologyanalysisandtheinternalprocesssimulation.Withboththeactivetestingandthepassivetesting,theroutingprotocoltestisgoingfurtherandmorethoroughlyandhelpsalotinthedevelopmentofrouters.
简介:许多软件系统在很多个连续版本被开发。在每个版本不仅新代码被增加而且存在代码经常被修改。在这研究,我们证明修改代码能是差错的重要来源。差错广泛地在软件项目作为主要费用司机之一被认出。因此,我们寻找在修改代码改进故障检测的方法。我们建议并且评估很多个预测模型增加故障检测的效率。为了造并且评估我们的模型,我们使用从Ericsson生产的二个大远程通信系统收集的数据。我们由比那两个都把他们用于系统的一个不同版本评估我们的模型的表演他们被造在上并且到一个不同系统。我们的模型的性能与理论最好的模型,的性能相比一个简单模型基于尺寸,以及到在一份随机的订单分析代码(不用任何模型)。我们发现我们的模型的使用在根本并且过去不基于班尺寸用一个简单模型使用任何模型上提供重要改进。我们的模型提供的获得对应于3857%理论最大增益。电子增补材料这篇文章(doi:10.1007/s11390-007-9053-3)的联机版本contatins增补材料,它对授权用户可得到。
简介:NondeterminismofPROLOGexecutionrequiresthatablockofcontrolinformationorachoicepointforeachprocedurecallbestoredwhenthereareothercandidateclausestobeused.Whenthecurrentlyselectedclausefails,thebindingsmadebytheclausemustbeundoneandthestoredchoicepointisreactivated,andthenanotherclauseofthecandidateonesischosentorunonit.StoringandreactivatingchoicepointsandundoingaccountforthegreatoverheadarerequiredtocontrolPROLOGexecution,whichisquitedifferentfromconventionalprograms.ThispaperfocusesonthetechniquesusedinSequentialPROLOGEngine(SPE)toreducetheoverheadofcontroloperations.ThecontrolinstructionsofSPEstorenomorechoicepointsthanthenecessary.Itsarchitecturetakestheapproachesofanalysingthepotentialparallelisminthecontroloperationsanddevelopingafractionofitduetothecost-effectconsideration.TheresultsofexecutingtwosampleprogramsonSPEintheformofhandtimingsarepresented,whichfavortheapproach.
简介:图象compositing广泛地被用来从分开的来源图象把视觉元素合为一幅单个图象。尽管最近的图象compositing技术能够完成从不同来源的视觉元素的光滑的相配,他们中的大多数含蓄地假设来源图象在一样的观点被拿。在这份报纸,我们在场从多重来源的compositing小说图象目标的一条途径想象哪个有不同观点。我们的关键想法是为来源图象目标的有意义的部件构造3D代理,并且使用这些3D部件代理在一样的观点一起弄歪并且无缝地合并部件。认识到这个想法,我们介绍一个并列框架的基于的单个看法的照相机刻度算法处理图象对象的一般类型,得到长方体的一个结构知道的长方体优化算法正确结构关系地为图象对象部件代理,并且最后,3D代理转变指导了图象变弯缝对象部件的算法。我们进一步基于这条compositing途径描述一个新奇应用程序自动地综合从一套模范的很多图象对象。试验性的结果证明我们的compositing途径能被用于许多图象目标,例如椅子,杯,灯,和机器人,并且合成申请能从模范的一个小集合与重要形状和风格变化创造新奇图象目标。
简介:Asteadyincreaseinconsumerdemands,andsevereconstraintsfrombothasomewhatdamagedenvironmentandnewlyinstalledgovernmentpolicies,requiretoday'sproductdesignanddevelopmenttobefasterandmoreefficientthaneverbefore,yetutilizingevenfewerresources.Newholisticapproaches,suchastotalproductlifecyclemodelingwhichembracesallaspectsofaproduct'slifecycle,arecurrentattemptstosolvetheseproblems.Withinthefieldofproductdesignandmodeling,featuretechnologyhasprovedtobeoneverypromisingsolutioncomponent.Owingtothetremendousincreaseininformationtechnology,totransferfromlowleveldataprocessingtowardsknowledgemodelingandinformationprocessingisabouttobringachangeinalmosteverycomputerizedapplication.Fromthisviewpoint,currentproblemsofbothfeatureframeworksandfeaturesystemsareanalyzedinrespecttostaticanddynamicconsistencybreakdowns.Theanalysisrangesfromearlystagesofdesigning(feature)conceptstofinalsystemimplementationandapplication.Forthefirsttime,anintegratedviewisgivenonapproaches,solutionsandpracticalexperience,withfeatureconceptsandstructures,providingbothafeatureframeworkanditsimplementationwithsufficientsystemarchitectureandcomputationalpowertomasterafairnumberofknownconsistencybreakdowns,whileprovidingforrobustcontextsforfeaturesemanticsandintegratedmodels.Withintoday'sheavyuseofinformationtechnologythesearepre-requisitesifthefullpotentialoffeaturetechnologyistobesuccessfullytranslatedintopractice.
简介:Sequentialpatternminingisanimportantdataminingproblemwithbroadapplications.However,itisalsoachallengingproblemsincetheminingmayhavetogenerateorexamineacombinatoriallyexplosivenumberofintermediatesubsequences.Recentstudieshavedevelopedtwomajorclassesofsequentialpatternminingmethods:(1)acandidategeneration-and-testapproach,representedby(i)GSP,ahorizontalformat-basedsequentialpatternminingmethod,and(ii)SPADE,averticalformat-basedmethod;and(2)apattern-growthmethod,representedbyPrefixSpananditsfurtherextensions,suchasgSpanforminingstructuredpatterns.Inthisstudy,weperformasystematicintroductionandpresentationofthepattern-growthmethodologyandstudyitsprinciplesandextensions.Wefirstintroducetwointerestingpattern-growthalgorithms,FreeSpanandPrefixSpan,forefficientsequentialpatternmining.ThenweintroducegSpanforminingstructuredpatternsusingthesamemethodology.Theirrelativeperformanceinlargedatabasesispresentedandanalyzed.Severalextensionsofthesemethodsarealsodiscussedinthepaper,includingminingmulti-level,multi-dimensionalpatternsandminingconstraint-basedpatterns.
简介:Modernsoftwaresystemsaresubjecttoacontinuousevolutionunderfrequentlyvaryingrequirementsandchangesinsystems'operationalenvironments.Lehman'slawofcontinuingchangedemandsforlong-livingandcontinuouslyevolvingsoftwaretoprolongitsproductivelifeandeconomicvaluebyaccommodatingchangesinexistingsoftware.Reusableknowledgeandpracticeshaveproventobesuccessfulforcontinuousdevelopmentandevolutionofthesoftwareeffectivelyandefficiently.However,challengessuchasempiricalacquisitionandsystematicapplicationofthereusableknowledgeandpracticesmustbeaddressedtoenableorenhancesoftwareevolution.Weinvestigatearchitecturechangelogs--mininghistoriesofarchitecture-centricsoftwareevolution--todiscoverchangepatternsthat1)supportreusabilityofarchitecturalchangesand2)enhancetheefficiencyofthearchitectureevolutionprocess.Wemodelarchitecturechangelogsasagraphandapplygraph-basedformalism(i.e.,graphminingtechniques)todiscoversoftwarearchitecturechangepatterns.Wehavedevelopedaprototypethatenablestool-drivenautomationanduserdecisionsupportduringsoftwareevolution.WehaveusedtheISO-IEC-9126modeltoqualitativelyevaluatetheproposedsolution.Theevaluationresultssuggestthattheproposedsolution1)enablesthereusabilityoffrequentarchitecturalchangesand2)enhancestheefficiencyofarchitecture-centricsoftwareevolutionprocess.Theproposedsolutionpromotesresearcheffortstoexploitthehistoryofarchitecturalchangestoempiricallydiscoverknowledgethatcanguidearchitecture-centricsoftwareevolution.
简介:Deduplication通常在两个企业存储系统和云存储被使用了。克服性能挑战为选择恢复deduplication系统的操作,solid-state-drive-based(即,基于SSD)读的缓存能为由缓冲加快被部署流行动态地恢复内容。不幸地,经常的数据更改由古典缓存计划导致了(例如,LRU和LFU)显著地弄短SSD一生当在SSD减慢I/O进程时。处理这个问题,我们建议新解决方案砍缓存极大地由扩大比例象I/O性能一样改进SSD的write耐久性长期流行(砍)在写进基于SSD的缓存的数据之中的数据。砍缓存保留很长时间在SSD缓存砍数据减少的时期缓存代替的数字。而且,它在deduplication集装箱阻止不得人心或不必要的数据被写进SSD缓存。我们在一个原型deduplication系统实现了砍缓存评估它的性能。我们的试验性的结果显示砍缓存弄短潜伏选择与仅仅deduplicated数据的5.56%能力以小基于SSD的缓存的成本由37.3%的一般水准恢复。重要地,砍缓存由9.77的一个因素改进SSD一生。砍缓存为一个成本效率的基于SSD的读的缓存解决方案提供到的证据表演增加性能选择为deduplication恢复系统。
简介:Inthefieldofmedicalimaging,thetraditionallocalbinarypattern(LBP)anditsimprovedalgorithmsareoftensensitivetonoise.TraditionalLBPsaresolelybasedonthesignalinformationfromlocaldifferences,andthebinaryquantizationmethodoversimplifiesthelocaltexturefeatureswhiledisregardingtheimaginginformationfromtheconcaveconvexregionsbetweenthehigh-orderpixelsandtheneighboringsamplingpoints.Therefore,weproposeanimprovedDerivedMeanCompleteLocalBinaryPattern(DM_CLBP)algorithmbasedonhigh-orderderivatives.IntheDM_CLBPmethod,thegreyvalueofasinglepixelisreplacedbythemeangreyvalueoftherectangularareablock,andthedifferencebetweenpixelvaluesintheareaisobtainedusingthesecond-orderdifferentiationmethod.Basedonthecalculationconceptofthecompletelocalbinarypattern(CLBP)algorithm,thecascadesignsandmagnitudesofthetwocomponentsareencodedandrecombinedinDM_CLBPusingauniformpattern.TheresultsfromtheexperimentsshowedthattheproposedDM_CLBPdescriptorsachievedaclassificationaccuracyof94.4%.ComparedwithLBPandotherimprovedalgorithms,theDM_CLBPalgorithmpresentedinthisstudycaneffectivelydifferentiatebetweenlesionareasandnormalareasinthyroidMR,(magneticresonance)imagesandshowstheimprovedaccuracyofareaclassification.