简介:与exascale来超级计算的时代,电源效率成为了最重要的障碍造一个exascale系统。Dataflow建筑学在为科学应用完成高电源效率有本国的优点。然而,最先进的dataflow体系结构没能为循环处理利用高并行。处理这个问题,我们建议一个pipelining环优化方法(PLO),它在处理元素(PE)在环流动做重复dataflow的数组加速器。这个方法由二种技术,帮助建筑学的硬件重复和帮助说明的软件重复组成。在硬件重复执行模型,一个在薄片上循环控制器被设计产生循环索引,减少计算内核并且打为pipelining执行的一个好基础的复杂性。在软件重复实行模型,另外的环指令被论述解决重复相关性问题。经由这二种技术,准备好了每周期执行的指令的平均数字被增加使浮点联合起来忙。当这二种技术的硬件费用是可接受的时,模拟结果证明分别地,我们的建议方法平均由2.45x和1.1x在浮点效率超过静电干扰和动态循环执行模型。
简介:Dempster-Shaferevidencetheory(DStheory)iswidelyusedinbrainmagneticresonanceimaging(MRI)segmentation,duetoitsefficientcombinationoftheevidencefromdifferentsources.Inthispaper,animprovedMRIsegmentationmethod,whichisbasedonfuzzyc-means(FCM)andDStheory,isproposed.Firstly,theaveragefusionmethodisusedtoreducetheuncertaintyandtheconflictinformationinthepictures.Then,theneighborhoodinformationandthedifferentinfluencesofspatiallocationofneighborhoodpixelsaretakenintoconsiderationtohandlethespatialinformation.Finally,thesegmentationandthesensordatafusionareachievedbyusingtheDStheory.ThesimulatedimagesandtheMRIimagesillustratethatourproposedmethodismoreeffectiveinimagesegmentation.
简介:Howtoletthespecialstubborndestructivecultbelieversrelievethedestructivecult’sspiritualcontrolandtobreakawayfromthedestructivecultisamajorproblemfacedbypeopleallovertheworldwhoaredevotingthemselvestotreatingthedestructivecultbelievers.Experienceshowsthattheuseofthe“latentlearningmethod”canachievegoodresults.Thearticleexpoundsthreepsychologicaltheoriesbasedonthislaw,includesthebehavioristpsychologist,Tolman’slatentlearningtheory,Bandura’sindirectlearningtheory,andphysiologicalpsychologistPenfield’ssurgicaldiscoverythatallthingsperceivedinthebrainarestoredinthebrain.Thisarticlealsointroducesthespecificoperationandapplicationofla
简介:Thecapabilityandreliabilityarecrucialcharacteristicsofmobilerobotswhilenavigatingincomplexenvironments.Theserobotsareexpectedtoperformmanyusefultaskswhichcanimprovethequalityoflifegreatly.Robotlocalizationanddecisionmakingarethemostimportantcognitiveprocessesduringnavigation.However,mostofthesealgorithmsarenotefficientandarechallengingtaskswhilerobotsnavigatethroughcomplexenvironments.Inthispaper,weproposeabiologicallyinspiredmethodforrobotdecision-making,basedonrat’sbrainsignals.Rodentsaccuratelyandrapidlynavigateincomplexspacesbylocalizingthemselvesinreferencetothesurroundingenvironmentallandmarks.Firstly,weanalyzedtherats’strategieswhilenavigatinginthecomplexY-maze,recordedlocalfieldpotentials(LFPs),simultaneously.TherecordedLFPswereprocessedanddifferentfeatureswereextractedwhichwereusedastheinputintheartificialneuralnetwork(ANN)topredicttherat’sdecision-makingineachjunction.TheANNperformancewastestedinarealrobotandgoodperformanceisachieved.Theimplementationofourmethodonarealrobot,demonstratesitsabilitiestoimitatetherat’sdecision-makingandintegratetheinternalstateswithexternalsensors,inordertoperformreliablenavigationincomplexmaze.
简介:Anovelefficienttrackinitiationmethodisproposedfortheharshunderwatertargettrackingenvironment(heavyclutterandlargemeasurementerrors):tracksplitting,evaluating,pruningandmergingmethod(TSEPM).Trackinitiationdemandsthatthemethodshoulddeterminetheexistenceandinitialstateofatargetquicklyandcorrectly.Heavyclutterandlargemeasurementerrorscertainlyposeadditionaldifficultiesandchallenges,whichdeteriorateandcomplicatethetrackinitiationintheharshunderwatertargettrackingenvironment.Therearethreeprimaryshortcomingsforthecurrenttrackinitiationmethodstoinitializeatarget:(a)theycannoteliminatetheturbulencesofcluttereffectively;(b)theremaybeahighfalsealarmprobabilityandlowdetectionprobabilityofatrack;(c)theycannotestimatetheinitialstateforanewconfirmedtrackcorrectly.Basedonthemultiplehypothesestrackingprincipleandmodifiedlogic-basedtrackinitiationmethod,inordertoincreasethedetectionprobabilityofatrack,tracksplittingcreatesalargenumberoftrackswhichincludethetruetrackoriginatedfromthetarget.Andinordertodecreasethefalsealarmprobability,basedontheevaluationmechanism,trackpruningandtrackmergingareproposedtoreducethefalsetracks.TSEPMmethodcandealwiththetrackinitiationproblemsderivedfromheavyclutterandlargemeasurementerrors,determinethetarget’sexistenceandestimateitsinitialstatewiththeleastsquaresmethod.What'smore,ourmethodisfullyautomaticanddoesnotrequireanykindmanualinputforinitializingandtuninganyparameter.Simulationresultsindicatethatournewmethodimprovessignificantlytheperformanceofthetrackinitiationintheharshunderwatertargettrackingenvironment.
简介:Inordertoachievetheinformationfusioninthetimedomainbasedontheevidencetheory,anevidencecombinationmethodinthetimedomainbasedonreliabilityandimportanceisproposedaccordingtotheideaofevidencediscount.Firstly,thedistortionofthetime-domainevidenceisjudgedbasedonsingleexponentialsmoothing.Thereal-timereliabilityoftheevidenceattheadjacenttimeisobtainedbythereal-timereliabilityassessmentmethodoftheevidencebasedonthecredibilitydecaymodel.Then,therelativeimportanceoftheevidenceattheadjacenttimeisobtainedbycomprehensivelyconsideringimprovedconflictdegreeanduncertainty.Finally,basedonthecriterionofevidencediscountandtheDempster’sruleofcombination,theevidencecombinationiscarriedouttoachievethesequentialcombinationoftime-domainevidence.Thenumericalsimulationandanalysisshowthatthismethodhasfullyembodiedthedynamiccharacteristicsoftime-domainevidencecombination,andithasstrongprocessingabilityforconflictinformationandanti-disturbingability.Theproposedmethodhasgoodapplicabilitytoinformationfusioninthetimedomain.
简介:ThispaperdevelopsanewlowerboundmethodforPOMDPsthatapproximatestheupdateofabeliefbytheupdateofitsnon-zerostates.ItusestheunderlyingMDPtoexploretheoptimalreachablestatespacefrominitialbeliefandselectactionsduringvalueiterations,whichsignificantlyacceleratestheconvergencespeed.Also,analgorithmwhichcollectsandprunesbeliefpointsbasedontheupperandlowerboundsispresented,andexperimentalresultsshowthatitoutperformssomeofthestate-of-artpoint-basedalgorithms.
简介:
简介:Streamprocessingapplicationscontinuouslyprocesslargeamountsofonlinestreamingdatainrealtimeornearrealtime.Theyhavestrictlatencyconstraints.However,thecontinuousprocessingmakesthemvulnerabletoanyfailures,andtherecoveriesmayslowdowntheentireprocessingpipelineandbreaklatencyconstraints.Theupstreambackupschemeisoneofthemostwidelyappliedfault-tolerantschemesforstreamprocessingsystems.Itintroducescomplexbackupdependenciestotasks,whichincreasesthedifficultyofcontrollingrecoverylatencies.Moreover,whendependenttasksarelocatedonthesameprocessor,theyfailatthesametimeinprocessor-levelfailures,bringingextrarecoverylatenciesthatincreasetheimpactsoffailures.Thispaperstudiestherelationshipbetweenthetaskallocationandtherecoverylatencyofastreamprocessingapplication.Wepresentacorrelatedfailureeffectmodeltodescribetherecoverylatencyofastreamtopologyinprocessor-levelfailuresunderataskallocationplan.Weintroducearecovery-latencyawaretaskallocationproblem(RTAP)thatseekstaskallocationplansforstreamtopologiesthatwillachieveguaranteedrecoverylatencies.WediscussthedifferencebetweenRTAPandclassictaskallocationproblemsandpresentaheuristicalgorithmwithacomputationalcomplexityofO(nlog2n)tosolvetheproblem.Extensiveexperimentswereconductedtoverifythecorrectnessandeffectivenessofourapproach.Itimprovestheresourceusageby15%-20%onaverage.
简介:Accordingtothepreciseephemerishasonlyprovidedsatellitepositionthatisdiscretenotanytime,soproposethatmakeuseofinterpolationmethodtocalculatesatellitepositionatanytime.TheessaytakeadvantageofIGSpreciseephemerisdatatocalculatesatellitepositionatsometimebyusingLagrangeinterpolation,Newtoninterpolation,Hermiteinterpolation,Cubicsplineinterpolationmethod,Chebyshevfittingmethodrespectively,whichhasadeeplyanalysisintheprecisionoffiveinterpolations.TheresultsshowthattheprecisionofCubicsplineinterpolationmethodistheworst,theprecisionofChebyshevfittingisbetterthanHermiteinterpolationmethod.LagrangeinterpolationandNewtoninterpolationarebetterthanothermethodsinprecision.Newtoninterpolationmethodhastheadvantagesofhighspeedandhighprecision.Therefore,Newtoninterpolationmethodhasacertainscientificsignificanceandpracticalvaluetogetthepositionofthesatellitequicklyandaccurately.
简介:Thebasicrequirementsof'businessprocesswalkthrough'methodwereclarifiedandthediagnosticwayofprocessproblemofenterprisewasdiscussed.Thebusinessprocessproblemscanbesummarizedasprocessstructuredefects,lackofsupportingmechanismsandlackofsupportingsystem.Thebusinessurgency-expectedreturnmatrixscoringmethod,whichcaneffectivelysorttheimportanceoftheprocessproblemswasalsoanalyzed.Theimplementationmethodandkeyelementsofprocessoptimizationwerediscussed,andtheevaluationindexsystemofprocessoptimizationwasalsoconstructed.Thecontinuouscustomer-orientedbusinessprocessoptimizationcaneffectivelyimprovethequalityoftheprocessoperationandenhancethelevelofenterpriseoperationmanagement.