简介:文[1]中提出了求解连续函数f(x)总体极小值的均值算法,并证明了算法的全局收敛性.若假设f(x)是定义在某可测集G上的可测函数,本文证明了均值算法产生的迭代序列全局收敛到f(x)的本质极小值,若进一步假设函数f(x)满足测度Lipschitz条件,还证明了求可测函数的均值算法是线性收敛的.
简介:通过构造拟上下解的单调迭代过程,在拟解对之间利用Sadvoskii不动点定理获得了Banach空间非线性三阶三点边值问题解的存在性.
简介:本文提出了求解非线性方程组的一种非精确Broyden方法.该方法是文献[8]中精确Broyden方法的推广.在适当的条件下,我们证明了非精确Broyden方法具有全局收敛性和超线性收敛性.数值实验表明,该方法效果较好.
简介:AinteriorpointscalingprojectedreducedHessianmethodwithcombinationofnonmonotonicbacktrackingtechniqueandtrustregionstrategyfornonlinearequalityconstrainedoptimizationwithnonegativeconstraintonvariablesisproposed.Inordertodealwithlargeproblems,apairoftrustregionsubproblemsinhorizontalandverticalsubspacesisusedtoreplacethegeneralfulltrustregionsubproblem.Thehorizontaltrustregionsubprobleminthealgorithmisonlyageneraltrustregionsubproblemwhiletheverticaltrustregionsubproblemisdefinedbyaparametersizeoftheverticaldirectionsubjectonlytoanellipsoidalconstraint.Bothtrustregionstrategyandlinesearchtechniqueateachiterationswitchtoobtainingabacktrackingstepgeneratedbythetwotrustregionsubproblems.Byadoptingthel1penaltyfunctionasthemeritfunction,theglobalconvergenceandfastlocalconvergencerateoftheproposedalgorithmareestablishedundersomereasonableconditions.AnonmonotoniccriterionandthesecondordercorrectionstepareusedtoovercomeMaratoseffectandspeeduptheconvergenceprogressinsomeill-conditionedcases.
简介:Thispaperdiscussestheintervalestimationsmethodfortheparametersandotherreliabilitycharactersofathree-poxameterWeibulldistribution.Accordingtothefiducialdistrlbutiontheoryoftheparameter,theauthorpresentstheconfidenceintervalsoftheporameters,thereliabilityandthereliablelife.Anexamplemadsimulationresultsaregiven.Itisshownthatthemethodpresentedinthispaperispracticableandworthnoticing.
简介:本文致力于研究非线性中立型延迟积分微分方程隐式Euler方法的收缩性。本文中的Lipschitz数是关于变量t的函数,而不是常数,最终能得到其数值解的结果是收缩的。