简介:本文利用Lyapunov—Krasovskii稳定性定理和线性矩阵不等式技术,得到了多时滞区间神经网路全局鲁棒稳定性的一个新的稳定性规则。该规则推广了最近文献中的一些结果,并通过数值仿真证实了结果的正确性。
简介:Afamilyofpiecewiserationalquinticinterpolationispresented.Eachinterpolationofthefamily,whichisidentifieduniquelybythevalueofaparameterαi,isofC2continuitywithoutsolvingasystemofconsistencyequationsforthederivativevaluesattheknots,andcanbeexpressedbythebasisfunctions.InterpolantisofO(hr)accuracywhenf(x)?Cr[a,b],andtheerrorshaveonlyasmallfloatingforabigchangeoftheparameterαi,itmeanstheinterpolationisstablefortheparameter.Theinterpolationcanpreservetheshapepropertiesofthegivendata,suchasmonotonicityandconvexity,andaproperchoiceofparameterαiisgiven.
简介:Resortingtocubicsplinefunctioninsteadofparametricsplinerepresentation,anexplicitfairnessindicatorandanefficientfairingalgorithmfor2Dcurvesarepresented.Theinputpointsequenceisfirstlypartitionedintoseveraloverlappedfunctionsegments.Foreachsegment,acubicsplinefunctionisusedastherepresentationtoolwhichentailsapolylineapproximationofcurvatureplot.Basedontheextrinsicrelationshipbetweenthepolylineandthepositionsofdatapoints,acoarse-to-finefaringmethodisproposedwhichefficientlyidentifiesandeliminatestheunnecessaryinflectionpoints.Ouralgorithmgeneratesthebestresultstodate,whichisvalidatedbynumerouspracticalexamples.