简介：Cytokineslikeinterferons(IFNs)playacentralroleinregulatinginnateandspecificimmunitiesagainstthepathogensandneoplasticcells.AnumberofsignalingpathwaysareinducedinresponsetoIFNinvariouscells.OneclassicmechanismemployedbyIFNsistheJAK-STATsignalingpathwayforinducingcellularresponses.Herewedescribethenon-STATpathwaysthatparticipateinIFN-inducedresponses.Inparticular,wewillfocusontheroleplayedbytranscriptionfactorC/EBP-βinmediatingtheseresponses.

简介：

简介：

简介：

简介：Asystemofthree-unitnetworkswithcoupledcellsisinvestigated.ThegeneralformulaforbifurcationdirectionofHopfbifurcationiscalculatedandtheestimateformulaofperiodoftheperiodicsolutionisgiven.

简介：胎儿的ECG抽取为胎儿的监视有重要意义。这份报纸介绍基于适应线性神经网络提取胎儿的ECG的一个方法。方法能被认识到bytraining数据的小数量。另外，更好的结果能被改进神经网络结构完成。因此，更容易鉴别胎儿的ECG能被提取。试验性的结果证明适应线性神经网络能被用来有效地从母亲的腹的信号提取胎儿的ECG。而且，更清楚的胎儿的ECG能被改进神经网络结构提取。

简介：Ａｎａｐｐｒｏａｃｈｏｆａｒｔｉｆｉｃｉａｌｎｅｕｒａｌｎｅｔｗｏｒｋｔｏｒｏｂｏｔｉｃｗｅｌｄｉｎｇｐｒｏｃｅｓｓｍｏｄｅｌｌｉｎｇ￥ＬｉＹａｎ（ＨａｒｂｉｎＲｅｓｅａｒｃｈＩｎｓｔｉｔｕｔｅｏｆＷｅｌｄｉｎｇＪｏｈｎＮｏｒｒｉｓｈａｎｄＴ．Ｅ．Ｂ...

简介：为多信道的脑电图(EEG)的独立来源分离(ISS)的一个神经网络方法在这份报纸被建议。用小浪的降噪的功能多尺度的分解，高周波的噪音从原版被移开(未加工)EEG。当加权的混合物和重量向量的表示被寻求第四顺序的cumulants的本地极值获得，然后，多信道的EEG被对待(即峭度系数)混合物。在这些进程步以后，加权的混合物被用作神经网络的输入，因此EEG的独立来源能一个一个地被分开。试验性的结果证明我们的方法为多信道的EEG的ISS是有效的。

简介：Toachievelowerassemblydelayatopticalburstswitchingedgenode,thispaperproposesanapproachcalledcurrentweightlengthprediction(CWLP)toimproveexistingestimatemechanisminburstassembly.isintroducedtomakeadynamictradeoffbetweenthecurrentandpasttrafficunderdifferentoffsettime.SimulationresultsshowthatCWLPcanachieveasignificantimprovementintermsoftrafficestimationinvariousoffsettimeandofferedload.

简介：

简介：Inthisstudy,aMulti-LayerBPneuralnetwork(MLBP)withdynamicthresholdsisemployedtobuildaclassifiermodel.Astothedesignoftheneuralnetworkstructure,theoreticalguidanceandplentifulexperimentsarecombinedtooptimizethehiddenlayers'parameterswhichincludethenumberofhiddenlayersandtheirnodenumbers.Theclassifierwithdynamicthresholdsisusedtostandardizetheoutputforthefirsttime,anditimprovestherobustnessofthemodeltoahighlevel.Finally,theclassifierisappliedtoforecastboxofficerevenueofamoviebeforeitstheatricalrelease.ThecomparisonresultswiththeMLPmethodshowthattheMLBPclassifiermodelachievesmoresatisfactoryresults,anditismorereliableandeffectivetosolvetheproblem.

简介：Intheenvironmentofheterogeneouswirelessnetworks,itisvitaltoselectacurrentlyoptimalnetworkforapplicationsandsubscribers.Theuseofmultipleattributedecisionmaking(MADM)forheterogeneousnetworkselectioncanprovidesubscriberswithsatisfactoryservicequality.ConvertingheterogeneousnetworkselectionintoaMADMproblem,theauthorspresentanimprovedalgorithmforMADMbasedongroupdecisiontheory.Thealgorithmcombinesweightvectorsofmultipleattributedecisionmakingtoobtainacombinationalweightvector.Thentheresults’compatibilitywillbeassessed.Iftheydonotmeettherequirementsofcompatibility,thejudgmentmatrixwillbemodifieduntilacomprehensivevectorthatsatisfiescompatibilityrequirementsisproduced.Thevectoriscombinedwithsimpleweightingmethod(SAW)fornetworkselection.Simulationshowsthatthealgorithmcanprovideuserswithsatisfactoryqualityofservice(QoS).

简介：

简介：BPandRBFneuralnetworktopredictforeststockvolumewerestudied,butthestudyinevaluatingbothnetworks’applicationeffectswasnotconducted.Inordertofindahigherforecastprecision,morestrongapplicativemethod,thecomprehensiveanalysisandevaluationonthetwomethodswerecarriedoutinthepracticalapplication.Bythecorrelationanalysis,crowndensity,shady-slopeandsunny-slope,TM1,TM2,TM3,TM5,TM7,NDVI,TM,(4-3),TM4/3wereselectedasinputvariables,andtheforestvolumeofMiyunCountyasoutputvariables,RBFandBPneuralnetworkmodelsforforecastingtheforestvolumewereestablished.Andtheneuralnetworktrainingsteplength,trainingtime,predictionaccuracyandtheapplicabilitymodelofthetwomethodswerecomprehensivelyanalyzed.TheresultsshowthattheRBFneuralnetworkmodelissuperiortotheBPneuralnetworkmodel.

简介：

简介：

简介：Afuzzyneuralnetworkcontrollerforunderwatervehicleshasmanyparametersdifficulttotunemanually.Toreducethenumerousworkandsubjectiveuncertaintiesinmanualadjustments,ahybridparticleswarmoptimization(HPSO)algorithmbasedonimmunetheoryandnonlineardecreasinginertiaweight(NDIW)strategyisproposed.OwingtotherestraintfactorandNDIWstrategy,anHPSOalgorithmcaneffectivelypreventprematureconvergenceandkeepbalancebetweenglobalandlocalsearchingabilities.Meanwhile,thealgorithmmaintainstheabilityofhandlingmultimodalandmultidimensionalproblems.TheHPSOalgorithmhasthefastestconvergencevelocityandfindsthebestsolutionscomparedtoGA,IGA,andbasicPSOalgorithminsimulationexperiments.ExperimentalresultsontheAUVsimulationplatformshowthatHPSO-basedcontrollersperformwellandhavestrongabilitiesagainstcurrentdisturbance.ItcanthusbeconcludedthattheproposedalgorithmisfeasibleforapplicationtoAUVs.

简介：

简介：Amethodologyisproposedtohandleproblemthatunderequiprobleaddressofpackettrafficattheinputport,GeneralizedShuffle-ExchangeNetwork(GSEN)routestrafficunevenlybecauseoftheunbalancedroutingtags.Theideaistouseroutingtagaccordingtoprobability,whichcanbeevaluatedbyusingMoore-Penroseinverseinmatrixanalysis.Aninstanceisusedtoillustratetheidea,andthesimulationisdonetoshowtheimprovementinperformanceissues.

简介：ThecharacteristicofEDFAinWDMsystemisanalyzed.ThesimulationsystembasedonthearchitectureoftheChinaInformationNetwork(CAINONET)isintroduced,andtheinfluenceofnoisefigure,gainunflatnessandoutputpowerofEDFAontheBERperformanceandscaleofnetworkisinvestigated.Moreover,theinfluenceofnoisefigureisstudiedwithdifferentcrosstalk.Meanwhile,itisindicatedthatitisimportanttoapplythe"nodeisolationprinciple"intheopticalnetwork.Theresearchresultscanprovidethevaluablereferenceforthepracticalconstructionofallopticalnetwork.