简介:Inthispaper,theL1-normestimatorsandtherandomweightedstatisticforasemiparametricregressionmodelareconstructed,thestrongconvergenceratesofestimatorsareobtainundercertainconditions,thestrongefficiencyoftherandomweightingmethodisshown.AsimulationstudyisconductedtocomparetheL1-normestimatorwiththeleastsquareestimatorintermofapproximateaccuracy,andsimulationresultsaregivenforcomparisonbetweentherandomweightingmethodandnormalapproximationmethod.
简介:Motiondeblurringisabasicprobleminthefieldofimageprocessingandanalysis.Thispaperproposesanewmethodofsingleimageblinddeblurringwhichcanbesignificanttokernelestimationandnon-blinddeconvolution.Experimentsshowthatthedetailsoftheimagedestroythestructureofthekernel,especiallywhentheblurkernelislarge.SoweextracttheimagestructurewithsalientedgesbythemethodbasedonRTV.Inaddition,thetraditionalmethodformotionblurkernelestimationbasedonsparsepriorsisconducivetogainasparseblurkernel.Butthesepriorsdonotensurethecontinuityofblurkernelandsometimesinducenoisyestimatedresults.ThereforeweproposethekernelrefinementmethodbasedonL0toovercometheaboveshortcomings.Intermsofnon-blinddeconvolutionweadopttheL1/L2regularizationterm.Comparedwiththetraditionalmethod,themethodbasedonL1/L2normhasbetteradaptabilitytoimagestructure,andtheconstructedenergyfunctionalcanbetterdescribethesharpimage.Forthismodel,aneffectivealgorithmispresentedbasedonalternatingminimizationalgorithm.
简介:在这份报纸,我们提供印射的Karush-Kuhn-Tucker(KKT)答案的柔韧的孤立的平静的完全的描述因为凸的抑制优化问题由原子标准功能调整了。这研究被最近的工作激发在[8],在作者在一个本地最佳的答案在Robinson限制资格下面显示出那的地方,为圆锥形的编程问题的一个宽类印射的KKT解决方案要用体力地被孤立如果资格(SRCQ)满足并且仅当两个都第二订足够的状况(SOSC)和严格的Robinson限制,安静。基于原子标准功能和它的conjugate的变化性质,我们建立在最初/双的SOSC和双/最初的SRCQ之间的等价。导出的结果导致印射的KKT答案的柔韧的孤立的平静的几相等的描述并且在原子标准的稳定性上把卓见加到存在文学调整凸的优化问题。
简介:Awell-knownresultforVilenkinsystemsisthefactthatforall1
norm.Thisstatementcannotbegeneralizedtoanyrepresentativeproductsystemonthecompleteproductoffinitenon-abeliangroups,buteventhenitistrueforthecompleteproductofquaterniongroupswithboundedordersandmonomialrepresentativeproductsystemorderedinaspecificway.
简介:双犹豫的模糊集合(DHFS)是由二部分组成的模糊集合(FS)的新归纳(即,会员迟疑功能和非会员迟疑工作),它面对显示认识的度的几不同可能的价值是否必然或无常。它包含模糊集合(FS),intuitionistic模糊集合(IFS),和犹豫的模糊集合(HFS)以便它能在决策的过程更灵活地处理不明确的信息。在这份报纸,我们基于爱因斯坦t-conorm和t标准在双犹豫的模糊集合上建议一些新操作,学习他们的性质和关系然后给一些双犹豫的模糊聚集操作员,它能被看作一些存在的归纳在下面模糊,intuitionistic模糊、犹豫的模糊环境。最后,在双犹豫的模糊环境下面的一个决策算法基于建议聚集操作员被给,一个数字例子被用来表明方法的有效性。
简介:Basedontherangespaceproperty(RSP),theequivalentconditionsbetweennonnegativesolutionstothepartialsparseandthecorrespondingweightedl_1-normminimizationproblemarestudiedinthispaper.Differentfromotherconditionsbasedonthesparkproperty,themutualcoherence,thenullspaceproperty(NSP)andtherestrictedisometryproperty(RIP),theRSPbasedconditionsareeasiertobeverified.Moreover,theproposedconditionsguaranteenotonlythestrongequivalence,butalsotheequivalencebetweenthetwoproblems.First,accordingtothefoundationofthestrictcomplementaritytheoremoflinearprogramming,asufficientandnecessarycondition,satisfyingtheRSPofthesensingmatrixandthefullcolumnrankpropertyofthecorrespondingsub-matrix,ispresentedfortheuniquenonnegativesolutiontotheweightedl_1-normminimizationproblem.Then,basedonthiscondition,theequivalenceconditionsbetweenthetwoproblemsareproposed.Finally,thispapershowsthatthematrixwiththeRSPoforderkcanguaranteethestrongequivalenceofthetwoproblems.
简介:ArobustdecentralizedH∞controlproblemforuncertainmulti-channelsystemsisconsidered.Theuncertaintiesareassumedtobetime-invariant,norm-bounded,andexistinboththesystemandcontrolinputmatrices.Thedynamicoutputfeedbackismainlydealtwith.Anecessaryandsufficientconditionfortheuncertainmulti-channelsystemtobestabilizedrobustlywithaspecifieddisturbanceattenuationlevelisderivedbasedontheboundedreallemma,whichisreducedtoafeasibilityproblemofanonlinearmatrixinequality(NMI).Atwo-stagehomotopymethodisusedtosolvetheNMIiteratively.First,adecentralizedcontrollerforthenominalsystemwithnouncertaintyiscomputedbyimposingstructuralconstraintsonthecoefficientmatricesofthecontrollergradually.Thenthedecentralizedcontrollerismodified,againgradually,tocopewiththeuncertainties.Oneachstage,avariableisfixedalternatelyattheiterationstoreducetheNMItoalinearmatrixinequality(LMI).Agivenexampleshowstheefficiencyofthismethod.