Optimality criteria without constraint qualifications for linear semidefinite problems

We consider two closely related optimization problems: a problem of convex semi-infinite programming with multidimensional index set and a linear problem of semi-definite programming. In the study of these problems we apply the approach suggested in our recent paper [14] and based on the notions of immobile indices and their immobility orders. For the linear semi-definite problem, we define the subspace of immobile indices and formulate the first-order optimality conditions in terms of a basic matrix of this subspace. These conditions are explicit, do not use constraint qualifications, and have the form of a criterion. An algorithm determining a basis of the subspace of immobile indices in a finite number of steps is suggested. The optimality conditions obtained are compared with other known optimality conditions.     

On vector and matrix median computation

The aim of this paper is to gain more insight into vector and matrix medians and to investigate algorithms to compute them. We prove relations between vector and matrix means and medians, particularly regarding the classical structure tensor. Moreover, we examine matrix medians corresponding to different unitarily invariant matrix norms for the case of symmetric 2×2 matrices, which frequently arise in image processing. Our findings are explained and illustrated by numerical examples. To solve the corresponding minimization problems, we propose several algorithms. Existing approaches include Weiszfeld’s algorithm for the computation of ?₂ vector medians and semi-definite programming, in particular, second order cone programming, which has been used for matrix median computation. In this paper, we adapt Weiszfeld’s algorithm for our setting and show that also two splitting methods, namely the alternating direction method of multipliers and the parallel proximal algorithm, can be applied for generalized vector and matrix median computations. Besides, we compare the performance of these algorithms numerically and apply them within local median filters.     

Duality for semi-definite and semi-infinite programming

In this article, we study semi-definite and semi-infinite programming problems (SDSIP), which includes semi-infinite linear programs and semi-definite programs as special cases. We establish that a uniform duality between the homogeneous (SDSIP) and its Lagrangian-type dual problem is equivalent to the closedness condition of certain cone. Moreover, this closedness condition was assured by a generalized canonically closedness condition and a Slater condition. Corresponding results for the nonhomogeneous (SDSIP) problem were obtained by transforming it into an equivalent homogeneous (SDSIP) problem.     

A cutting plane algorithm for semi-definite programming problems with applications to failure discriminant analysis

We will propose a new cutting plane algorithm for solving a class of semi-definite programming problems (SDP) with a small number of variables and a large number of constraints. Problems of this type appear when we try to classify a large number of multi-dimensional data into two groups by a hyper-ellipsoidal surface. Among such examples are cancer diagnosis, failure discrimination of enterprises. Also, a certain class of option pricing problems can be formulated as this type of problem. We will show that the cutting plane algorithm is much more efficient than the standard interior point algorithms for solving SDP.     

二次半定规划问题及其投影收缩算法

In this paper,we discuss the relations among the quadratic semi-definite programming problem,the linear semi-definite porgramming and the linearquadratic semi-definite programming problem.The duality theories are presented.After proving the equivalence of its optimality conditions and monotonous linear variational inequalities,we use the projection and contraction algorithms to solve(QSDP),We present the algorithms and its convergence analysis.     

Strong duality for inexact linear programming

In this paper, we apply the Dubovitskii-Milyutin approach to derive strong duality theorems for inexact linear programming problems. Inexact linear programming deals with the standard linear problem in which the data is not well known and it is supposed to lie in certain given convex sets. The case of parametric dependence of the data is particularly analyzed and relations with semi-infinite and

semi-definite programming are also commented.     

模糊多目标半定规划

This paper first applies the fuzzy set theory to multi-objective semi-definite program-ming （MSDP）, and proposes the fuzzy multi-objective semi-definite programming （FMSDP） model whose optimal efficient solution is defined for the first time, too. By constructing a membership function, the FMSDP is translated to the MSDP. Then we prove that the optimal efficient solution of FMSDP is consistent with the efficient solution of MSDP and present the optimality condition about these programming. At last, we give an algorithm for FMSDP by introducing a new membership function and a series of transformation.     

AnUV-algorithm for semi-infinite multiobjective programming

In this paper, we give an application ofUV-decomposition method of convex programming to multiobjective programming, and offer a new algorithm for solving semi-infinite multiobjective programming. Finally, the superlinear convergence of the algorithm is proved.     

On numerical optimization theory of infinite kernel learning

In Machine Learning algorithms, one of the crucial issues is the representation of the data. As the given data source become heterogeneous and the data are large-scale, multiple kernel methods help to classify “nonlinear data”. Nevertheless, the finite combinations of kernels are limited up to a finite choice. In order to overcome this discrepancy, a novel method of  “infinite”  kernel combinations is proposed with the help of infinite and semi-infinite programming regarding all elements in kernel space. Looking at all infinitesimally fine convex combinations of the kernels from the infinite kernel set, the margin is maximized subject to an infinite number of constraints with a compact index set and an additional (Riemann–Stieltjes) integral constraint due to the combinations. After a parametrization in the space of probability measures, it becomes semi-infinite. We adapt well-known numerical methods to our infinite kernel learning model and analyze the existence of solutions and convergence for the given algorithms. We implement our new algorithm called “infinite” kernel learning (IKL) on heterogenous data sets by using exchange method and conceptual reduction method, which are well known numerical techniques from solve semi-infinite programming. The results show that our IKL approach improves the classifaction accuracy efficiently on heterogeneous data compared to classical one-kernel approaches.     

The Simplex and the Dual Method for Quadratic Programming

The paper presents a straightforward generalization of the Simplex and the dual method for linear programming to the case of convex quadratic programming. The two algorithms, called the Simplex and the dual method for quadratic programming, are applicable when the matrix of the quadratic part of the objective function, in case this function is to be maximized, is negative definite, negative semi-definite or zero; in the last case the two methods are equivalent to an application of the similar methods for linear programming. The paper gives an exposition of the methods as well as examples and interpretations. The relations with linear programming methods are considered and some starting procedures in case no initial feasible solution is available are presented.     

半无限规划离散化问题一个两阶段序列二次规划算法

本文研究了求解半无限规划离散化问题(P)的一个新的算法.利用序列二次规划(SQP)两阶段方法和约束指标集的修正技术,提出了求解(P)的一个两阶段SQP算法.算法结构简单,搜索方向的计算成本较低.在适当的条件下,证明了算法具有全局收敛性.数值试验结果表明算法是有效的.推广了文献[4]中求解(P)的算法.     

On the Lipschitz Modulus of the Argmin Mapping in Linear Semi-Infinite Optimization

This paper is devoted to quantify the Lipschitzian behavior of the optimal solutions set in linear optimization under perturbations of the objective function and the right hand side of the constraints (inequalities). In our model, the set indexing the constraints is assumed to be a compact metric space and all coefficients depend continuously on the index. The paper provides a lower bound on the Lipschitz modulus of the optimal set mapping (also called argmin mapping), which, under our assumptions, is single-valued and Lipschitz continuous near the nominal parameter. This lower bound turns out to be the exact modulus in ordinary linear programming, as well as in the semi-infinite case under some additional hypothesis which always holds for dimensions n ? 3. The expression for the lower bound (or exact modulus) only depends on the nominal problem’s coefficients, providing an operative formula from the practical side, specially in the particular framework of ordinary linear programming, where it constitutes the sharp Lipschitz constant. In the semi-infinite case, the problem of whether or not the lower bound equals the exact modulus for n > 3 under weaker hypotheses (or none) remains as an open problem.     

An evaluation of mathematical programming and minicomputers

The availability of efficient mathematical software on minicomputers could greatly increase the use of operations research techniques in industry and government. The objective of this paper is to demonstrate the feasibility of implementing a particular class of mathematical programming algorithms, namely shortest path algorithms, on “typical” minicomputers. Two distinct shortest path algorithms were tested on four computer systems using a common set of test problems. Computational results are presented which verify the feasibility of implementing these algorithms in a minicomputer environment, and also show the relative efficiency of each algorithm to be the same when tested on a minicomputer as when tested on a large-scale computer system.     

Posynomial geometric programming as a special case of semi-infinite linear programming

This paper develops a wholly linear formulation of the posynomial geometric programming problem. It is shown that the primal geometric programming problem is equivalent to a semi-infinite linear program, and the dual problem is equivalent to a generalized linear program. Furthermore, the duality results that are available for the traditionally defined primal-dual pair are readily obtained from the duality theory for semi-infinite linear programs. It is also shown that two efficient algorithms (one primal based and the other dual based) for geometric programming actually operate on the semi-infinite linear program and its dual.     

A parameterized hessian quadratic programming problem

We present a general active set algorithm for the solution of a convex quadratic programming problem having a parametrized Hessian matrix. The parametric Hessian matrix is a positive semidefinite Hessian matrix plus a real parameter multiplying a symmetric matrix of rank one or two. The algorithm solves the problem for all parameter values in the open interval upon which the parametric Hessian is positive semidefinite. The algorithm is general in that any of several existing quadratic programming algorithms can be extended in a straightforward manner for the solution of the parametric Hessian problem.This research was supported by the Natural Sciences and Engineering Research Council under Grant No. A8189 and under a Postgraduate Scholarship, by an Ontario Graduate Scholarship, and by the University of Windsor Research Board under Grant No. 9432.     

A parameterized hessian quadratic programming problem

We present a general active set algorithm for the solution of a convex quadratic programming problem having a parametrized Hessian matrix. The parametric Hessian matrix is a positive semidefinite Hessian matrix plus a real parameter multiplying a symmetric matrix of rank one or two. The algorithm solves the problem for all parameter values in the open interval upon which the parametric Hessian is positive semidefinite. The algorithm is general in that any of several existing quadratic programming algorithms can be extended in a straightforward manner for the solution of the parametric Hessian problem. This research was supported by the Natural Sciences and Engineering Research Council under Grant No. A8189 and under a Postgraduate Scholarship, by an Ontario Graduate Scholarship, and by the University of Windsor Research Board under Grant No. 9432.     

On level sets of marginal functions

The investigation of level sets of marginal functions is motivated by several aspects of standard and generalized semi-infinite programming. The feasible- set M of such a prob­lem is easily seen to be a level set of the marginal function corresponding to the lower level problem. In the present paper we study the local structure of M at feasible bound­ary points in the generic case. A codimension formula shows that there is a wide range of these generic situations, but that the number of active indices is always bounded by the state space dimension. We restrict our attention to two special subcases

In the first case, where the number of active indices is maximalM is shown to be locally diffeomorphic to the non-negative orthant. This situation is well-known from finite and also from standard semi-infinite programming. However, in the second case a generic situation arises which is typical for generalized semi-infinite programming. Here, the active index set is a singleton, and M can exhibit a re-entrant corner or even local non-closedness, depending on whether the Mangasarian-Fromovitz constraint qualifica­tion holds at the active index. If an objective function is minimized over M then in the setting of the second case a local minimizer cannot occur     

An accelerated central cutting plane algorithm for linear semi-infinite programming

An algorithm for linear semi-infinite programming is presented which accelerates the convergence of the central cutting plane algorithm first proposed in [4]. Compared with other algorithms, the algorithm in [4] has the advantage of being applicable under mild conditions and of providing feasible solutions. However its convergence has been shown to be rather slow in practical instances. The algorithm proposed in this paper introduces a simple acceleration scheme which gives faster convergence, as confirmed by several examples, as well as an interval of prefixed length containing the optimum value. It is also shown that the algorithm provides a solution of the dual problem and that it can be used for convex semi-infinite programming too.Mathematics Subject Classification (1991): 90C05, 90C34, 65K05, 90C51Acknowledgments. The author whishes to thank the three anonymous referees and an associate editor for many useful comments and valuable suggestions.