简介:研究了具时变时滞的分层抑制细胞神经网络.利用不动点定理获得了若干判定该网络存在概周期解的新充分条件,改进和推广了已有文献中的相应结论.
简介:AinteriorpointscalingprojectedreducedHessianmethodwithcombinationofnonmonotonicbacktrackingtechniqueandtrustregionstrategyfornonlinearequalityconstrainedoptimizationwithnonegativeconstraintonvariablesisproposed.Inordertodealwithlargeproblems,apairoftrustregionsubproblemsinhorizontalandverticalsubspacesisusedtoreplacethegeneralfulltrustregionsubproblem.Thehorizontaltrustregionsubprobleminthealgorithmisonlyageneraltrustregionsubproblemwhiletheverticaltrustregionsubproblemisdefinedbyaparametersizeoftheverticaldirectionsubjectonlytoanellipsoidalconstraint.Bothtrustregionstrategyandlinesearchtechniqueateachiterationswitchtoobtainingabacktrackingstepgeneratedbythetwotrustregionsubproblems.Byadoptingthel1penaltyfunctionasthemeritfunction,theglobalconvergenceandfastlocalconvergencerateoftheproposedalgorithmareestablishedundersomereasonableconditions.AnonmonotoniccriterionandthesecondordercorrectionstepareusedtoovercomeMaratoseffectandspeeduptheconvergenceprogressinsomeill-conditionedcases.
简介:基于Chen-Harker-Kanzow-Smale光滑函数,对单调非线性互补问题NCP(f)给出了一种不可行非内点连续算法,该算法在每次迭代时只需求解一个线性等式系统,执行一次线搜索;算法在NCP(f)的解处不需要严格互补的条件下,具有全局线性收敛性和局部二次收敛性.
简介:线性矩阵不等式的优良性质可用于解决细胞神经网络中的保性能控制问题.本文介绍了线性矩阵不等式的相关概念和性质;通过对Schur补引理的改进提出了一个引理,从而更容易将二次矩阵不等式转化为线性矩阵不等式,更好地应用于控制参数求解;提出了LMI的基本问题和MATLAB工具箱,并对LMI在细胞神经网络的保性能控制问题作出了简要描述.