Team selection for prediction tasks

Given a random variable \(O \in \mathbb {R}\) and a set of experts \(E\), we describe a method for finding a subset of experts \(S \subseteq E\) whose aggregated opinion best predicts the outcome of \(O\). Therefore, the problem can be regarded as a team formation for performing a prediction task. We show that in case of aggregating experts’ opinions by simple averaging, finding the best team (the team with the lowest total error during past \(k\) rounds) can be modeled with an integer quadratic programming and we prove its NP-hardness whereas its relaxation is solvable in polynomial time. At the end, we do an experimental comparison between different rounding and greedy heuristics on artificial datasets which are generated based on calibration and informativeness of exprets’ information and show that our suggested tabu search works effectively.     

Exact solution method to solve large scale integer quadratic multidimensional knapsack problems

In this paper we develop a branch-and-bound algorithm for solving a particular integer quadratic multi-knapsack problem. The problem we study is defined as the maximization of a concave separable quadratic objective function over a convex set of linear constraints and bounded integer variables. Our exact solution method is based on the computation of an upper bound and also includes pre-procedure techniques in order to reduce the problem size before starting the branch-and-bound process. We lead a numerical comparison between our method and three other existing algorithms. The approach we propose outperforms other procedures for large-scaled instances (up to 2000 variables and constraints). A extended abstract of this paper appeared in LNCS 4362, pp. 456–464, 2007.     

Optimization techniques for multivariate least trimmed absolute deviation estimation

Given a dataset an outlier can be defined as an observation that does not follow the statistical properties of the majority of the data. Computation of the location estimate is of fundamental importance in data analysis, and it is well known in statistics that classical methods, such as taking the sample average, can be greatly affected by the presence of outliers in the data. Using the median instead of the mean can partially resolve this issue but not completely. For the univariate case, a robust version of the median is the Least Trimmed Absolute Deviation (LTAD) robust estimator introduced in Tableman (Stat Probab Lett 19(5):387–398, 1994), which has desirable asymptotic properties such as robustness, consistently, high breakdown and normality. There are different generalizations of the LTAD for multivariate data, depending on the choice of norm. Chatzinakos et al. (J Comb Optim, 2015) we present such a generalization using the Euclidean norm and propose a solution technique for the resulting combinatorial optimization problem, based on a necessary condition, that results in a highly convergent local search algorithm. In this subsequent work, we use the \(L^1\) norm to generalize the LTAD to higher dimensions, and show that the resulting mixed integer programming problem has an integral relaxation, after applying an appropriate data transformation. Moreover, we utilize the structure of the problem to show that the resulting LP’s can be solved efficiently using a subgradient optimization approach. The robust statistical properties of the proposed estimator are verified by extensive computational results.     

On a general framework for network representability in discrete optimization

In discrete optimization, representing an objective function as an s-t cut function of a network is a basic technique to design an efficient minimization algorithm. A network representable function can be minimized by computing a minimum s-t cut of a directed network, which is an efficiently solvable problem. Hence it is natural to ask what functions are network representable. In the case of pseudo Boolean functions (functions on \(\{0,1\}^n\)), it is known that any submodular function on \(\{0,1\}^3\) is network representable. ?ivný–Cohen–Jeavons showed by using the theory of expressive power that a certain submodular function on \(\{0,1\}^4\) is not network representable. In this paper, we introduce a general framework for the network representability of functions on \(D^n\), where D is an arbitrary finite set. We completely characterize network representable functions on \(\{0,1\}^n\) in our new definition. We can apply the expressive power theory to the network representability in the proposed definition. We prove that some ternary bisubmodular function and some binary k-submodular function are not network representable.     

Tracking and outperforming large stock-market indices

Enhanced index-tracking funds aim to achieve a small target excess return over a given financial benchmark index with minimum additional risk relative to this index, i.e., a minimum tracking error. These funds are attractive to investors, especially when the index is large and thus well diversified. We consider the problem of determining a portfolio for an enhanced index-tracking fund that is benchmarked against a large stock-market index subject to real-life constraints that may be imposed by investors, stock exchanges, or investment guidelines. In the literature, various solution approaches have been proposed to enhanced index tracking that are based on different linear and quadratic tracking-error functions. However, it remains an open question which tracking-error function should be minimized to determine good enhanced index-tracking portfolios. Moreover, the existing approaches may neglect real-life constraints such as the minimum trading values imposed by stock exchanges or may not devise good feasible portfolios within a reasonable computational time when the index is large. To overcome these shortcomings, we propose novel mixed-integer linear and quadratic programming formulations and novel matheuristics. To address the open question, we minimize different tracking-error functions by applying the proposed matheuristics and exact solution approaches based on the proposed mixed-integer programming formulations in a computational experiment using a set of problem instances based on large stock-market indices with up to more than 9,000 constituents. The results of our study suggest that minimizing the so-called tracking error variance, which is a quadratic function, is preferable to minimizing other tracking-error functions.     

Optimization techniques for robust multivariate location and scatter estimation

Computation of typical statistical sample estimates such as the median or least squares fit usually require the solution of an unconstrained optimization problem with a convex objective function, that can be solved efficiently by various methods. The presence of outliers in the data dictates the computation of a robust estimate, which can be defined as the optimum statistical estimate for a subset that contains at least half of the observations. The resulting problem is now a combinatorial optimization problem which is often computationally intractable. Classical statistical methods for multivariate location \(\varvec{\mu }\) and scatter matrix \(\varvec{\varSigma }\) estimation are based on the sample mean vector and covariance matrix, which are very sensitive in the presence of outlier observations. We propose a new method for robust location and scatter estimation which is composed of two stages. In the first stage an unbiased multivariate \(L_{1}\)-median center for all the observations is attained by a novel procedure called the least trimmed Euclidean deviations estimator. This robust median defines a coverage set of observations which is used in the second stage to iteratively compute the set of outliers which violate the correlational structure of the data set. Extensive computational experiments indicate that the proposed method outperforms existing methods in accuracy, robustness and computational time.     

A Branch and Bound algorithm for general mixed-integer quadratic programs based on quadratic convex relaxation

Let \((MQP)\) be a general mixed-integer quadratic program that consists of minimizing a quadratic function \(f(x) = x^TQx +c^Tx\) subject to linear constraints. Our approach to solve \((MQP)\) is first to consider an equivalent general mixed-integer quadratic problem. This equivalent problem has additional variables \(y_{ij}\) , additional quadratic constraints \(y_{ij}=x_ix_j\) , a convex objective function, and a set of valid inequalities. Contrarily to the reformulation proposed in Billionnet et al. (Math Program 131(1):381–401, 2012), the equivalent problem cannot be directly solved by a standard solver. Here, we propose a new Branch and Bound process based on the relaxation of the non-convex constraints \(y_{ij}=x_ix_j\) to solve \((MQP)\) . Computational experiences are carried out on pure- and mixed-integer quadratic instances. The results show that the solution time of most of the considered instances with up to 60 variables is improved by our Branch and Bound algorithm in comparison with the approach of Billionnet et al. (2012) and with the general mixed-integer nonlinear solver BARON (Sahinidis and Tawarmalani, Global optimization of mixed-integer nonlinear programs, user’s manual, 2010).     

On nonlinear multi-covering problems

In this paper we define the exact k-coverage problem, and study it for the special cases of intervals and circular-arcs. Given a set system consisting of a ground set of n points with integer demands \(\{d_0,\dots ,d_{n-1}\}\) and integer rewards, subsets of points, and an integer k, select up to k subsets such that the sum of rewards of the covered points is maximized, where point i is covered if exactly \(d_i\) subsets containing it are selected. Here we study this problem and some related optimization problems. We prove that the exact k-coverage problem with unbounded demands is NP-hard even for intervals on the real line and unit rewards. Our NP-hardness proof uses instances where some of the natural parameters of the problem are unbounded (each of these parameters is linear in the number of points). We show that this property is essential, as if we restrict (at least) one of these parameters to be a constant, then the problem is polynomial time solvable. Our polynomial time algorithms are given for various generalizations of the problem (in the setting where one of the parameters is a constant).     

Base polyhedra and the linking property

An integer polyhedron \(P \subseteq {\mathbb {R}}^n\) has the linking property if for any \(f \in {\mathbb {Z}}^n\) and \(g \in {\mathbb {Z}}^n\) with \(f \le g\), P has an integer point between f and g if and only if it has both an integer point above f and an integer point below g. We prove that an integer polyhedron in the hyperplane \(\sum _{j=1}^n x_j=\beta \) is a base polyhedron if and only if it has the linking property. The result implies that an integer polyhedron has the strong linking property, as defined in Frank and Király (in: Cook, Lovász, Vygen (eds) Research trends in combinatorial optimization, Springer, Berlin, pp 87–126, 2009), if and only if it is a generalized polymatroid.     

Integer programming approach to static monopolies in graphs

A subset M of vertices of a graph is called a static monopoly, if any vertex v outside M has at least \(\lceil \tfrac{1 }{2}\deg (v)\rceil \) neighbors in M. The minimum static monopoly problem has been extensively studied in graph theoretical context. We study this problem from an integer programming point of view for the first time and give a linear formulation for it. We study the facial structure of the corresponding polytope, classify facet defining inequalities of the integer programming formulation and introduce some families of valid inequalities. We show that in the presence of a vertex cut or an edge cut in the graph, the problem can be solved more efficiently by adding some strong valid inequalities. An algorithm is given that solves the minimum monopoly problem in trees and cactus graphs in linear time. We test our methods by performing several experiments on randomly generated graphs. A software package is introduced that solves the minimum monopoly problem using open source integer linear programming solvers.     

Heuristics for the network design problem with connectivity requirements

We consider the NP-complete problem of finding a spanning \(k\)-tree of minimum weight in a complete weighted graph. This problem has a number of applications in designing reliable backbone telecommunication networks. We propose effective algorithms based on a greedy strategy and several variable neighborhood search metaheuristics. We also develop an integer linear programming model for calculating a lower bound. Preliminary numerical experiments using random and real-word data sets are reported to show the effectiveness of our approach. In addition, we compare our approach with known metaheuristics.     

A Combined D.C. Optimization—Ellipsoidal Branch-and-Bound Algorithm for Solving Nonconvex Quadratic Programming Problems

In this paper we propose a new branch-and-bound algorithm by using an ellipsoidal partition for minimizing an indefinite quadratic function over a bounded polyhedral convex set which is not necessarily given explicitly by a system of linear inequalities and/or equalities. It is required that for this set there exists an efficient algorithm to verify whether a point is feasible, and to find a violated constraint if this point is not feasible. The algorithm is based upon the fact that the problem of minimizing an indefinite quadratic form over an ellipsoid can be efficiently solved by some available (polynomial and nonpolynomial time) algorithms. In particular, the d.c. (difference of convex functions) algorithm (DCA) with restarting procedure recently introduced by Pham Dinh Tao and L.T. Hoai An is applied to globally solving this problem. DCA is also used for locally solving the nonconvex quadratic program. It is restarted with current best feasible points in the branch-and-bound scheme, and improved them in its turn. The combined DCA-ellipsoidal branch-and-bound algorithm then enhances the convergence: it reduces considerably the upper bound and thereby a lot of ellipsoids can be eliminated from further consideration. Several numerical experiments are given.     

Approximation Algorithms for Quadratic Programming

We consider the problem of approximating the global minimum of a general quadratic program (QP) with n variables subject to m ellipsoidal constraints. For m=1, we rigorously show that an -minimizer, where error (0, 1), can be obtained in polynomial time, meaning that the number of arithmetic operations is a polynomial in n, m, and log(1/). For m 2, we present a polynomial-time (1- )-approximation algorithm as well as a semidefinite programming relaxation for this problem. In addition, we present approximation algorithms for solving QP under the box constraints and the assignment polytope constraints.     

Nonparametric Predictive Inference for Exposure Assessment

Exposure assessment for food and drink consumption requires the combining of information about people's consumption of products with concentration data sets to provide predictions for chemical intake by humans. In this article, we present a method called nonparametric predictive inference (NPI) for exposure assessment. NPI is a distribution‐free method relying only on Hill's assumption . Effectively, is a postdata exchangeability assumption, which is a natural starting point for nonparametric statistics. For further discussion we refer to works by Hill and Coolen. We illustrate how NPI can be implemented to produce predictions for an individual's exposure based on consumption, body weight, and concentration data. NPI has the advantage that we do not have to assume a distribution to implement it. There may, however, be information available to suggest a distribution for a random quantity. Therefore, we present an NPI‐Bayes hybrid method where this information can be taken into account by using Bayesian methods while using NPI for the other random quantities in the model.     

Modified linear programming and class 0 bounds for graph pebbling

Given a configuration of pebbles on the vertices of a connected graph G, a pebbling move removes two pebbles from some vertex and places one pebble on an adjacent vertex. The pebbling number of a graph G is the smallest integer k such that for each vertex v and each configuration of k pebbles on G there is a sequence of pebbling moves that places at least one pebble on v. First, we improve on results of Hurlbert, who introduced a linear optimization technique for graph pebbling. In particular, we use a different set of weight functions, based on graphs more general than trees. We apply this new idea to some graphs from Hurlbert’s paper to give improved bounds on their pebbling numbers. Second, we investigate the structure of Class 0 graphs with few edges. We show that every n-vertex Class 0 graph has at least \(\frac{5}{3}n - \frac{11}{3}\) edges. This disproves a conjecture of Blasiak et al. For diameter 2 graphs, we strengthen this lower bound to \(2n - 5\), which is best possible. Further, we characterize the graphs where the bound holds with equality and extend the argument to obtain an identical bound for diameter 2 graphs with no cut-vertex.     

A continuous characterization of the maximum vertex-weighted clique in hypergraphs

For a simple graph G on n vertices with adjacency matrix A, Motzkin and Strauss established a remarkable connection between the clique number and the global maximum value of the quadratic programm: \(\textit{max}\{ \mathbf {x}^T A \mathbf {x}\}\) on the standard simplex: \(\{\sum _{i=1}^{n} x_i =1, x_i \ge 0 \}\). In Gibbons et al. (Math Oper Res 122:754–768, 1997), an extension of the Motzkin–Straus formulation was provided for the vertex-weighted clique number of a graph. In this paper, we provide a continuous characterization of the maximum vertex-weighted clique problem for vertex-weighted uniform hypergraphs.     

A new effective branch-and-bound algorithm to the high order MIMO detection problem

This paper develops a branch-and-bound method based on a new convex reformulation to solve the high order MIMO detection problem. First, we transform the original problem into a \(\{-1,1\}\) constrained quadratic programming problem with the smallest size. The size of the reformulated problem is smaller than those problems derived by some traditional transformation methods. Then, we propose a new convex reformulation which gets the maximized average objective value as the lower bound estimator in the branch-and-bound scheme. This estimator balances very well between effectiveness and computational cost. Thus, the branch-and-bound algorithm achieves a high total efficiency. Several simulations are used to compare the performances of our method and other benchmark methods. The results show that this proposed algorithm is very competitive for high accuracy and relatively good efficiency.     

An efficient algorithm for distance total domination in block graphs

The \(k\)-distance total domination problem is to find a minimum vertex set \(D\) of a graph such that every vertex of the graph is within distance \(k\) from some vertex of \(D\) other than itself, where \(k\) is a fixed positive integer. In the present paper, by using a labeling method, we design an efficient algorithm for solving the \(k\)-distance total domination problem on block graphs, a superclass of trees.     

基于条件概率积分变换的多元Copula函数选择

本文构建了基于条件概率积分变换的Copula函数选择方法,通过对条件概率积分变换下Anderson-Darling(AD)、Kolmogorov-Smirnov(KS)、Cramér-von Mises(CM)这三种统计量的比较,讨论在不同样本容量和变量维数下其对多种Copula函数的拟合效果。利用GSPTSE、INMEX.MX和NDX三大股指样本,将基于条件概率积分变换的Copula函数选择方法与核密度估计和极大似然估计选择法的效果进行系统比较。结果表明,基于条件概率积分变换的检验法可以有效解决多元Copula函数的选择问题,其拟合优度检验更精确、更稳定;核密度估计检验在大样本下比较稳定,而小样本下稳定性较差;相比之下,极大似然值检验法则不稳定。     

Parameterized dominating set problem in chordal graphs: complexity and lower bound

In this paper, we study the parameterized dominating set problem in chordal graphs. The goal of the problem is to determine whether a given chordal graph G=(V,E) contains a dominating set of size k or not, where k is an integer parameter. We show that the problem is W[1]-hard and it cannot be solved in time unless 3SAT can be solved in subexponential time. In addition, we show that the upper bound of this problem can be improved to when the underlying graph G is an interval graph.