Construction of best Bregman approximations in reflexive Banach spaces

An iterative method is proposed to construct the Bregman projection of a point onto a countable intersection of closed convex sets in a reflexive Banach space.

     

On the optimal design of the randomized unbiased Monte Carlo estimators

We consider a class of unbiased Monte Carlo estimators and develop an efficient algorithm to produce the distribution of the optimal random truncation level. We establish the convergence and optimality results of the associated algorithm and also derive its exact complexity. We find this algorithm has a much lower complexity as compared to the existing one in the literature. The proposed algorithm is also applicable to optimization problems arising in supply chain management, such as economic reorder interval problem.     

Estimation of a covariance matrix with location: Asymptotic formulas and optimal B-robust estimators

Applying the non-singular affine transformations AZ + μ to a spherically symmetrically distributed variate Z generates the covariance-location model, indexed by the parameters AA^T and μ, consisting of so-called elliptical distributions. We develop an algebraic machinery that simplifies the derivation of influence functions and asymptotic variance-covariance matrices for equivariant estimators of Σ and μ and reveals a natural structure of Σ. In addition, optimal B-robust estimators are derived.     

A randomized generalized low rank approximations of matrices algorithm for high dimensionality reduction and image compression

High‐dimensionality reduction techniques are very important tools in machine learning and data mining. The method of generalized low rank approximations of matrices (GLRAM) is a popular technique for dimensionality reduction and image compression. However, it suffers from heavily computational overhead in practice, especially for data with high dimension. In order to reduce the cost of this algorithm, we propose a randomized GLRAM algorithm based on randomized singular value decomposition (RSVD). The theoretical contribution of our work is threefold. First, we discuss the decaying property of singular values of the matrices during iterations of the GLRAM algorithm, and provide a target rank required in the RSVD process from a theoretical point of view. Second, we establish the relationship between the reconstruction errors generated by the standard GLRAM algorithm and the randomized GLRAM algorithm. It is shown that the reconstruction errors generated by the former and the latter are comparable, even if the solutions are computed inaccurately during iterations. Third, the convergence of the randomized GLRAM algorithm is investigated. Numerical experiments on some real‐world data sets illustrate the superiority of our proposed algorithm over its original counterpart and some state‐of‐the‐art GLRAM‐type algorithms.     

Characterization of best Chebyshev approximations with prescribed norm

In this paper a new characterization of smooth normed linear spaces is discussed using the notion of proximal points of a pair of convex sets. It is proved that a normed linear space is smooth if and only if for each pair of convex sets, points which are mutually nearest to each other from the respective sets are proximal.     

Constructing one-parameter families of elliptic curves with moderate rank

We give several new constructions for moderate rank elliptic curves over Q(T). In particular we construct infinitely many rational elliptic surfaces (not in Weierstrass form) of rank 6 over Q using polynomials of degree two in T. While our method generates linearly independent points, we are able to show the rank is exactly 6 without having to verify the points are independent. The method generalizes; however, the higher rank surfaces are not rational, and we need to check that the constructed points are linearly independent.     

GAMMA－MINIMAX ESTIMATORS FOR THE MEAN OF A MULTIVARIATE NORMAL DISTRIBUTION WITH PARTIALLY UNKNOWN COVARIANCE MATRIX

ＧＡＭＭＡ－ＭＩＮＩＭＡＸＥＳＴＩＭＡＴＯＲＳＦＯＲＴＨＥＭＥＡＮＯＦＡＭＵＬＴＩＶＡＲＩＡＴＥＮＯＲＭＡＬＤＩＳＴＲＩＢＵＴＩＯＮＷＩＴＨＰＡＲＴＩＡＬＬＹＵＮＫＮＯＷＮＣＯＶＡＲＩＡＮＣＥＭＡＴＲＩＸＣＨＥＮＬＡＮＸＩＮＧ（陈兰祥）（Ｄｅｐａｒｔ...     

Jordan canonical form of three-way tensor with multilinear rank (4,4,3)

The Jordan canonical form of matrix is an important concept in linear algebra. And the concept has been extended to three-way arrays case. In this paper, we study the Jordan canonical form of three-way tensor with multilinear rank (4,4,3). For a 4 × 4 × 4 tensor gj with multilinear rank (4,4,3), we show that gj must be turned into the canonical form if the upper triangular entries of the last three slices of gj are nonzero. If some of the upper triangular entries of the last three slices of gj are zeros, we give some conditions to guarantee that gj can be turned into the canonical form.     

Representations of best linear unbiased estimators in the Gauss-Markoff model with a singular dispersion matrix

In the general Gauss-Markoff model (Y, Xβ, σ²V), when V is singular, there exist linear functions of Y which vanish with probability 1 imposing some restrictions on Y as well as on the unknown β. In all earlier work on linear estimation, representations of best-linear unbiased estimators (BLUE's) are obtained under the assumption: “L′Y is unbiased for Xβ ? L′X = X.” Such a condition is not, however, necessary. The present paper provides all possible representations of the BLUE's some of which violate the condition L′X = X. Representations of X for given classes of BLUE's are also obtained.     

约束下线性混合模型参数估计的渐近解及其收敛性

在线性不等式约束下讨论了具有相同参数的两个线性混合模型的参数估计问题,给出了一种迭代算法,得到了这类模型中参数的最小二乘估计序列及其渐近解.在此基础上,利用多元多项式方程组解的个数定理和不动点定理,证明了此估计序列是依概率1收敛的.     

A numerical method for solving an optimal control problem with fixed endpoints of trajectories and a convex functional

We propose a numerical method for solving the problem of taking a nonlinear system to the zero state while minimizing a nonnegative convex functional. The particular cases include the problem of minimizing the consumption of resources or energy. The method is based on the maximum principle and approximations of bodies with some families of simplices. The properties of coverings of bodies with simplices enable us to justify the convergence of the method.     

A trust region method for minimization of nonsmooth functions with linear constraints

We introduce a trust region algorithm for minimization of nonsmooth functions with linear constraints. At each iteration, the objective function is approximated by a model function that satisfies a set of assumptions stated recently by Qi and Sun in the context of unconstrained nonsmooth optimization. The trust region iteration begins with the resolution of an “easy problem”, as in recent works of Martínez and Santos and Friedlander, Martínez and Santos, for smooth constrained optimization. In practical implementations we use the infinity norm for defining the trust region, which fits well with the domain of the problem. We prove global convergence and report numerical experiments related to a parameter estimation problem. Supported by FAPESP (Grant 90/3724-6), FINEP and FAEP-UNICAMP. Supported by FAPESP (Grant 90/3724-6 and grant 93/1515-9).     

A mortar edge element method with nearly optimal convergence for three-dimensional Maxwell's equations

In this paper, we are concerned with mortar edge element methods for solving three-dimensional Maxwell's equations. A new type of Lagrange multiplier space is introduced to impose the weak continuity of the tangential components of the edge element solutions across the interfaces between neighboring subdomains. The mortar edge element method is shown to have nearly optimal convergence under some natural regularity assumptions when nested triangulations are assumed on the interfaces. A generalized edge element interpolation is introduced which plays a crucial role in establishing the nearly optimal convergence. The theoretically predicted convergence is confirmed by numerical experiments.

     

Numbers with fixed sum of digits in linear recurrent number systems

We study the set of integers with a given sum of digits with respect to a linear recurrent digit system. An asymptotic formula for the number of integers ≤N with given sum of digits is determined, and the distribution in residue classes is investigated, thus generalizing results due to Mauduit and Sárközy. It turns out that numbers with fixed sum of digits are uniformly distributed in residue classes under some very general conditions. Namely, the underlying linear recurring sequence must have the property that there is no prime factor P of the modulus such that all but finitely many members of the sequence leave the same residue modulo P. The key step in the proof is an estimate for exponential sums using known theorems from Diophantine approximation.     

对数正态分布场合有非常数尺度参数恒加试验的参数估计

讨论了对数正态分布场合有非常数尺度参数恒加寿命试验的参数估计，导出了基于全样本和定数截尾样本恒加试验的点估计和近似区间估计。     

Newsvendor equations for optimal reorder levels of serial inventory systems with fixed batch sizes

We consider a stochastic serial inventory system with a given fixed batch size per stage and linear inventory holding and penalty costs. For this system, echelon stock (R,nQ) policies are known to be optimal. On the basis of new average costs formulas, we obtain newsvendor equations for the optimal reorder levels.     

Convergence criterion of Newton's method for singular systems with constant rank derivatives

The present paper is concerned with the convergence problem of Newton's method to solve singular systems of equations with constant rank derivatives. Under the hypothesis that the derivatives satisfy a type of weak Lipschitz condition, a convergence criterion based on the information around the initial point is established for Newton's method for singular systems of equations with constant rank derivatives. Applications to two special and important cases: the classical Lipschitz condition and the Smale's assumption, are provided; the latter, in particular, extends and improves the corresponding result due to Dedieu and Kim in [J.P. Dedieu, M. Kim, Newton's method for analytic systems of equations with constant rank derivatives, J. Complexity 18 (2002) 187-209].     

Analytical approximations to predict performance measures of markovian type manufacturing systems with job failures and parallel processing

Manufacturing or service systems with multiple product classes, job circulation due to random failures, resources shared between product classes, and some portions of the manufacturing or assembly carried in series and the rest in parallel are commonly observed in real-life. The web server assembly is one such manufacturing system which exhibits the above characteristics. Predicting the performance measures of these manufacturing systems is not an easy task. The primary objective of this research was to propose analytical approximations to predict the flow times of the manufacturing systems, with the above characteristics, and evaluate its accuracy. The manufacturing system is represented as a network of queues. The parametric decomposition approach is used to develop analytical approximations for a system with arrival and service rates from a Markovian distribution. The results from the analytical approximations are compared to simulation models. In order to bridge the gap in error, correction terms were developed through regression modeling. The experimental study conducted indicates that the analytical approximations along with the correction terms can serve as a good estimate for the flow times of the manufacturing systems with the above characteristics.