动态多人递阶决策问题——结构,非光滑性及必要条件

利用非光滑分析、集值映射和微分包含理论，分析了动态递阶决策问题的结构、性质，给出了解决这类问题的一个通用性结构化模型，并利用这一模型在非凸非光滑条件下导出了最优策略存在的必要条件     

非光滑两级优化问题的必要条件及其算法

利用集值分析、非光滑分析与优化理论,研究了两级优化问题的性质、结构,提出了处理这类问题的一个通用性结构化模型,并利用该模型导出了非凸非光滑条件下两级优化问题的微分包含型必要条件,最后给出了其求解的算法及算例．     

电力备用市场资源优化配置的模型与求解方法

对于发电公司在电力备用市场中的资源优化配置问题,建立一种均值一方差模型,同时允许发电公司从市场购买某种产品以满足其它产品的需求.针对模型中出现的非光滑问题,引入极大熵函数将其光滑化,并采用非线性互补方法求解.该方法利用Karush-Kuhn-Tuchker(KKT)条件得到一个非线性互补问题,然后利用非线性互补函数将其转化为非光滑方程组,并引入参数再次进行光滑化,最终通过求解一组光滑方程组以逼近原问题的最优解.数值分析验证了该方法的有效性.     

金融风险管理中ES度量的非参数方法的比较及其应用

预期不足(ES)是近几年发展起来的用于测量和控制金融风险的量化工具.在金融时间序列中, 将两步核估计应用于两步ES非参数估计之中,得到了ES模型的两步核光滑估计. 通过计算其期望和方差,比较了两步核光滑ES估计与ES完全经验估计及一步核光滑估计的优劣,得到了有趣的结论: 与VaR模型不同, 两步光滑化并不能减小ES估计的方差,反而会增大其方差, 并通过计算机模拟证实了理论获得的结论.对国内沪深两市中的封闭式基金进行了实证分析,计算了样本基金的ES完全经验估计、一步核光滑估计和两步核光滑估计,并计算了样本基金基于周收益率和ES的两步核光滑估计的风险调整收益(RAROC),以此对样本基金的业绩做出了评价. 实证分析表明: 在不同的置信水平下,基于周收益率和ES计算的风险调整收益排名比基于周收益率和VaR计算的风险调整收益排名要更加稳定.     

动态非光滑Nash协商解必要条件的研究

本文利用非光滑分析理论，采用公理化方法研究了非光滑情形下的动态协商对策问题，给出其Ｎａｓｈ协商解的必要条件，并给予了严格的数学证明．     

具有主从结构的非光滑两层优化问题

本文研究一类具有主从结构的非光滑两层化问题。基于Ｃｌａｒｋｅ意义下的非光滑分析理论，给出了该类问题的最优性条件，并对模型的求解方法进行了讨论。     

基于非光滑方程组的智能电网实时定价

通过研究智能电网实时定价问题,发现通过社会效益模型可以将短时段实时定价问题转化为一个互补问题.根据互补理论和经济学含义,互补问题又可以转化为包含影子价格的非光滑方程组.采用影子价格作为定价基础,通过构造光滑函数逼近非光滑方程组和拟牛顿法求解,得到基础电价.数值仿真表明,新方法下的社会效益与传统固定定价方法下的社会效益相近,但是新方法下的电价更低,计算速度快,收敛性好,数值结果稳定.     

多项式光滑的半监督支持向量分类机

为了处理半监督支持向量分类优化中的非凸非光滑问题,引入一族多项式光滑函数来逼近非凸的目标函数,给出的多项式函数在样本的高密度区逼近精度高,逼近精度低时出现在样本的低密度区,同时可以根据不同的精度要求选择不同的逼近函数.采用BFGS算法求解模型.在人工数据和UCI数据集上的实验结果显示,算法不仅能保证标号数据很少时的分类精度,而且不因标号数据的增多而明显提高分类性能,因此给出的分类器性能是稳定的.     

线性时变经济系统分支稳定性研究:生存理论方法

研究线性时变经济系统的分支稳定性问题 .定义了描述系统状态约束的集值映射 ,利用生存理论 ,分别给出了在状态约束为光滑和非光滑时系统分支渐近稳定的充要条件 .     

光滑支持向量机多项式函数的研究

为了找到多项式光滑支持向量机(polynomial smooth support vector machine,PSSVM)中性能更好的光滑函数,将正号函数变形并展开为多项式级数,得到一类光滑函数。证明了这类函数的性能,它既能满足任意阶光滑的要求,也能达到任意给定的逼近精度。用Newton-Armijo算法求解相应的PSSVM模型,实验结果表明,随着多项式光滑函数阶数的提高,逼近精度和相应PSSVM模型的分类性能也相应提高。     

APPROXIMATE SUBDIFFERENTIALS AND NONSMOOTH ANALYSIS FINITE DIMENSIONS

Ioffe's approximate subdifferentials are reviewed and some of his resultsare generalized.An extension of the calculus of the approximate subdifferentials forthe sums to any finite number of functions is provided along with a generalizationof the Dubovitzkii-Milyutin theorem.The presentation also indicates some of thelimitations of nonsmooth analysis and optimization.Restriction to the class offunction which is suitable for most of the purposes in nonsmooth optimization issuggested.     

Sufficiency and duality for nonsmooth multiobjective programming problems involving generalized univex functions

In this paper, nonsmooth univex, nonsmooth quasiunivex, and nonsmooth pseudounivex functions are introduced. By utilizing these new concepts, sufficient optimality conditions for a weakly efficient solution of the nonsmooth multiobjective programming problem are established. Weak and strong duality theorems axe also derived for Mond-Weir type multiobjective dual programs.     

Nonsmooth Equations of K-T Systems for a Constrained Minimax Problem

1 .INTRODUCTIONTheconstrainedminimaxproblemplaysanimportantroleinoptimizationsinceithasmanyapplicationsinengineering,systemanalysis,managementscienceandsoon .Therearemanypublicationsdealingwiththisproblem .Inthepresentpaper,weintendtoex plorethenonsmoothequationapproachtosolvetheconstrainedminimaxproblem (P2 )below .Thatis,wefirsttransformthisproblemintoanequivalentsystemofthenonsmoothequation ,thenusegeneral izedNewtonmethodstosolvethenonsmoothequa tion .Wenowconsiderthenonlinear progra…     

An Asset Allocation Model and Its Solving Method

Asset allocation is an important issue in finance, and both risk and return are its fundamental ingredients. Rather than the return, the measure of the risk is complicated and of controversy.In this paper, we propose an appropriate risk measure which is precisely a convex combination of mean semi-deviation and conditional value-at-risk. Based on this risk measure, investors can trade-off flexibly between the volatility and the loss to tackle the incurring risk by choosing different convex coefficients.As the presented risk measure contains nonsmooth term, the asset allocation model based on it is nonsmooth. To employ traditional gradient algorithms, we develop a uniform smooth approximation of the plus function and convert the model into a smooth one. Finally, an illustrative empirical study is given. The results indicate that investors can control risk efficiently by adjusting the convex coefficient and the confidence level simultaneously according to their perceptions. Moreover, the effectiveness of the smoothing function proposed in the paper is verified.     

Newton type methods for solving nonsmooth equations

1.INTRODUCTION Thenonsmoothequationsproblemplaysanimportant roleinengineering,systemanalysis,management scienceandetc.Manypublicationsdealtwiththe methodsforsolutionofnonsmoothequations,see Refs.[1～4].Letusconsiderthesystemofnons moothequations F(x)=0(1)whereF∶Rn→RmislocallyLipschitzian.NewtonmethodsforsolvingthenonsmoothEq.(1)aregiven asfollows xk+1=xk-V-1kF(xk)(2)whereVkisanelementofvarioussubdifferentialsof Fatxk,forinstance,Clarkesubdifferential[1],B differential[2],b differen…     

Delay differential equations under nonlinear impulsive control and applications to neural network models

In this paper,a class of delay differential equations with nonlinear impulsive control is discussed.Based on the nonsmooth analysis,criteria of stability are obtained for delay differential equations with nonlinear impulses control under certain conditions.These criteria can be applied to some neural network models.At the end of the paper,two examples are provided to illustrate the feasibility and effectiveness of the proposed results.