Epi-Convergence and Lower and Epi-Upper Semicontinuous Approximations in Distribution

The paper introduces an extension of the epi-convergence, the lower semicontinuous approximation and the epi-upper semicontinuous approximation of random real functions in distribution. The new notions could be helpful tools for sensitivity analyzes of stochastic optimization problems. The research is evoked by S. Vogel and continues the research started by Vogel and the author.     

Stability of multistage stochastic programming

In this paper, we study the stability of multistage stochastic programming with recourse in a way that is different from that used in studying stability of two-stage stochastic programs. Here, we transform the multistage programs into mathematical programs in the space ⁿ ×L _p with a simple objective function and multistage stochastic constraints. By investigating the continuity of the multistage multifunction defined by the multistage stochastic constraints and applying epi-convergence theory we obtain stability results for linear and linear-quadratic multistage stochastic programs.Project supported by the National Natural Science Foundation of China.     

Convergence results for a self-dual regularization of convex problems

We study a one-parameter regularization technique for convex optimization problems whose main feature is self-duality with respect to the Legendre–Fenchel conjugation. The self-dual technique, introduced by Goebel, can be defined for both convex and saddle functions. When applied to the latter, we show that if a saddle function has at least one saddle point, then the sequence of saddle points of the regularized saddle functions converges to the saddle point of minimal norm of the original one. For convex problems with inequality and state constraints, we apply the regularization directly on the objective and constraint functions, and show that, under suitable conditions, the associated Lagrangians of the regularized problem hypo/epi-converge to the original Lagrangian, and that the associated value functions also epi-converge to the original one. Finally, we find explicit conditions ensuring that the regularized sequence satisfies Slater's condition.     

随机规划中的一些逼近结果

主要讨论了一类随机规划的目标函数分别在概率测度序列分布收敛、函数序列上图收敛以及随机变量序列均方可积收敛等收敛意义下目标函数序列的收敛情况。基于上述收敛情况给出了一些逼近思想，这些思想可应用于求解这类随机规划问题。     

Limiting Distributions of Linear Programming Estimators

Smith (1994) proposes estimation in linear regression models with non-negative errors by maximizing the sum of fitted values subject to the constraint that the fitted values can be no larger than the corresponding response value. In this paper, we consider the limiting distribution of these estimators under very general conditions. Some extensions to local polynomial estimation are also considered.     

Approximations of Bayes decision problems: the epigraphical approach

Solving Bayesian decision problems usually requires approximation procedures, all leading to study the convergence of the approximating infima. This aspect is analysed in the context of epigraphical convergence of integral functionals, as minimal context for convergence of infima. The results, applied to the Monte Carlo importance sampling, give a necessary and sufficient condition for convergence of the approximations of Bayes decision problems and sufficient conditions for a large class of Bayesian statistical decision problems.     

A Lévy Type Martingale Convergence Theorem for Random Sets with Unbounded Values

Given a nondecreasing sequence ( _n ) of sub--fields and a real or vector valued random variable f, the Lévy Martingale convergence Theorem (LMCT) asserts that E(f/ _n ) converges to E(f/) almost surely and in L ¹, where stands for the -field generated by the _n . In the present paper, we study the validity of the multivalued analog this theorem for a random set F whose values are members of (X), the space of nonempty closed sets of a Banach space X, when (X) is endowed either with the Painlevé–Kuratowski convergence or its infinite dimensional extensions. We deduce epi-convergence results for integrands via the epigraphical multifunctions. As it is known, these results are useful for approximating optimization problems. The method relies on countability supportness hypotheses which are shown to hold when the values of the random set E(F/ _n ) do not contain any line. On the other hand, since the values of F are not assumed to be bounded, conditions involving barrier and asymptotic cones are shown to be necessary. Moreover, we discuss the relations with other multivalued martingale convergence theorems and provide examples showing the role of the hypotheses. Even in the finite dimensional setting, our results are new or subsume already existing ones.     

Stability of Slopes and Subdifferentials

We show that the slope introduced by DeGiorgi, Marino and Tosques in 1980 is stable (in a sense to be made precise) with respect to the variational convergence introduced by the authors in 2000. Applications to the stability of subdifferentials are derived, including a further characterization of slice-convergence of convex functions in terms of set convergence of enlargements of their subdifferentials.     

Penalization in non-classical convex programming via variational convergence

A penalty method for convex functions which cannot necessarily be extended outside their effective domains by an everywhere finite convex function is proposed and combined with the proximal method. Proofs of convergence rely on variational convergence theory.     

An example of stability for the minima of a sequence of dc functions: Homogenization for a class of nonlinear Sturm-Liouville problems

Combining the Clarke-Ekeland dual least action principle and the epi-convergence, we state an existence result and study the asymptotic behaviour for the periodic solution of a nonlinear Sturm-Liouville problem deriving from a convex subquadratic potential, when the data are perturbed in a suitable sense. The result appears like a stability result for the minimizers of a sequence of DC functions.