简介:Researchonjumpregressionfunc(?)onshasnotbeenadequateyetAccordingtotheinformationaboutthenumberofjumps,theirpositionsandjumpmagnitudes,jumpregressionfunctionscanbeclassifiedintoeighttypes.Thispaperdealsespeciallywiththesecondjumpregressionfunction.Firstofall,aconceptoftrimmedsplineestimateisproposedandwithitanL~2-consistentestimateofthesmoothingpartofthejumpregressionfunctionisobtained.ThisalongwiththeL~2-consistentestimateofjumpmagnitudeconstitutesanestimateofthesecondjumpregressionfunction.Thispaperdiscussesalsothecasethatthejumppositionshavesomeindeterminacy.Anewcriterionissuggestedanditsuniquesolutionderived.Intheend,afewnumericalresultsaregiven.
简介:上一讲中,我们借题发挥,谈及了内容与形式的重要性。这一讲中,我们亦借助这个题目,通过对这个问题的不同回答来深化一下这个讨论。从考试的情况来看,许多考生对这个问题都有自己的观点,有些还很有见解。然而,他们在表达中明显地受到了语言水平的限制,因而出现辞不达意、表达形式单调、不够自然流畅等问题。所以笔者针对考生在回答时暴露出的上
简介:
简介:Oneofthechiefmaterialsusedintheconstructionofbuildingsandroadsiseement(水泥),apowdermademainlyfromlimestone(石灰石)andclay(粘土).EventheancientEgyptiansandRomans(罗马人)usedakindofcement,butitwasnotuntil1824thatanEnglishbricklayer(泥瓦匠)developedacementstrongenoughformodemroadsandbuilding.
简介:Deadlockmustbeavoidedinamanufacturingsystem.Inthispaper,anefficientalgorithmforfindinganoptimaldeadlock-freeschedulesinamanufacturingsystemwithverylimitedbufferispresented.Thisalgorithmisbasedonheeffectivegeneticalgorithm(GA)searchmethod,andaformalPetrinetstructureisintroducedtodetectthetokenplayerassuringdeadlock-free.InordertomaketheschedulingstrategygeneratedbyGAmeettherequiredconstraintofdeadlock-free,someresultsofthestruetureanalysisofPetrinetareinvolvedasacriteriontoselectdeadlock-freeschedulefromthepopulationgeneratedbyGA.Theeffectivenessandefficiencyoftheproposedapproachisillustratedbyusinganexample.
简介:ThesufficientconditionsofHoldercontinuityoftwokindsoffractalinterpolationfunctionsdefinedbyIFS(IteratedFunctionSystem)wereobtained.Thesufficientandnecessaryconditionforitsdifferentiabilitywasproved.ItsderivativewasafractalinterpolationfunctiongeneratedbytheassociatedIFS,ifitisdifferentiable.
简介:WeproposethevariationaldescriptionofgeneratingfunctionapproachoffirstkindforHamil-tonianODEs,andextendtheapproachtothesemi-linearwaveequations.Inthisway,wecanconstructanyfiniteorderaccuracyscheme,andshowthattheresultingnumericalschemeismultisymplectic.Atlast,wepresentsomenumericalexperimentsbyusingderivednewscheme.
简介:Inthisnote,wedefineanewkindofspacesTFofTriebel-LizorkintypeandcharacterizeitviasequencespacelpAlsoaFouriermultipliertheoremisestablishedinthisnewspacesTFpaq
简介:TheauthorsfirstproveaconvergenceresultontheKacanovmethodforsolvinggeneralnonlineaxvafiationalinequalitiesofthesecondkindandthenapplytheKacanovmethodtosolveanonlinearvariationalinequalityofthesecondkindarisinginelastoplastieity.Inadditiontotheconvergenceresult,anaposteriorierrorestimateisshownfortheKacanoviterates.IneachstepoftheKacanoviteration,onehasa(linear)variationalinequalityofthesecondkind,whichcanhesolvedbyuaingaregularizationtechnique.TheKacanoviterationandtheregularizationtechniquetogetherprovideapproximationswhichcanbereadilycomputednumerically.AnaposteriorierrorestimateisderivedforthecombinedeffectoftheKacanoviterationandtheregularization.
简介:§1.IntroductionItisknownthatthefollowingCauchyproblemforaparabolicpartialdifferentialequation(wherethevaluesattherightboundary,u.(1,t)=v(t)areunknownandsoughtfor)isill-posed:thesolution(v)doesnotdependcontinuouslyonthedata(g).Inordertotreattheill-posednessanddevelopthenumericalmethod,onereformulatestheproblemasaVolterraintegralequationofthefirstkindwishaconvolutiontypekernel(seeSneddon[1],CarslawandJaeger[2])
简介:Conjugategradientmethods.areaclassofimportantmethodsforunconstrainedoptimization,especiallywhenthedimensionislarge.In2001,DaiandLiaohaveproposedanewconjugatecondition,basedonittwononlinearconjugategradientmethodsareconstructed.Withtrustregionidea,thispapergivesaself-adaptivetechniqueforthetwomethods.Thenumericalresultsshowthatthistechniqueworkswellforthegivennonlinearoptimizationtestproblems.
简介:Wepresentafastalgorithmbasedonpolynomialinterpolationtoapproximatematricesarisingfromthediscretizationofsecond-kindintegralequationswherethekernelfunctioniseithersmooth,non-oscillatoryandpossessingonlyafinitenumberofsingularitiesoraproductofsuchfunctionwithahighlyoscillatorycoefficientfunction.Contrasttowavelet-likeapproximations,ourapproximationmatrixisnotsparse.However,theapproximationcanbeconstrucedinO(n)operationsandrequiresO(n)storage,wherenisthenumberofquadraturepointsusedinthediscretization.Moreover,thematrix-vectormultiplicationcostisoforderO(nlogn).Thusourschemeiswellsuitableforconjugategradienttypemethods.Ournumericalresultsindicatethatthealgorithmisveryaccurateandstableforhighdegreepolynomialinterpolation.
简介:ThispaperdealswiththeapproximatesolutionoftheFredholmequationLu=fofthesecondkindwithf∈H’(Γ).First,wegivetheorderofthen-theoptimalradiusintheworstcasesetting.Then,weshowthattheGalerkinmethodusing2n+1innerproductsoffhasminimalerror.Finally,wegivetheestimateoftheε-complexityoftheFredholmproblemofthesecondkindandGalerkinmethodintheworstcasesetting.
简介:最近,Brutman和Passow认为Newman类型合理插值是to|x|由对称的节点的任意的集合导致了在[-1,1]并且给了近似错误的一般评价。由他们的方法,一个人能为一些特殊节点建立近似的准确顺序。在我们考虑插值节点是第二种类型的零个Chebyshev多项式并且在这种情况中证明那的特殊情况的现在的纸,近似的准确顺序是O(1/(nlnn))。