A sparse proximal implementation of the LP dual active set algorithm

We present an implementation of the LP Dual Active Set Algorithm (LP DASA) based on a quadratic proximal approximation, a strategy for dropping inactive equations from the constraints, and recently developed algorithms for updating a sparse Cholesky factorization after a low-rank change. Although our main focus is linear programming, the first and second-order proximal techniques that we develop are applicable to general concave–convex Lagrangians and to linear equality and inequality constraints. We use Netlib LP test problems to compare our proximal implementation of LP DASA to Simplex and Barrier algorithms as implemented in CPLEX. This material is based upon work supported by the National Science Foundation under Grant No. 0203270.     

Relaxation-based algorithms for minimax optimization problems with resource allocation applications

We consider minimax optimization problems where each term in the objective function is a continuous, strictly decreasing function of a single variable and the constraints are linear. We develop relaxation-based algorithms to solve such problems. At each iteration, a relaxed minimax problem is solved, providing either an optimal solution or a better lower bound. We develop a general methodology for such relaxation schemes for the minimax optimization problem. The feasibility tests and formulation of subsequent relaxed problems can be done by using Phase I of the Simplex method and the Farkas multipliers provided by the final Simplex tableau when the corresponding problem is infeasible. Such relaxation-based algorithms are particularly attractive when the minimax optimization problem exhibits additional structure. We explore special structures for which the relaxed problem is formulated as a minimax problem with knapsack type constraints; efficient algorithms exist to solve such problems. The relaxation schemes are also adapted to solve certain resource allocation problems with substitutable resources. There, instead of Phase I of the Simplex method, a max-flow algorithm is used to test feasibility and formulate new relaxed problems.Corresponding author.Work was partially done while visiting AT&T Bell Laboratories.     

A numerically stable form of the simplex algorithm

Standard implementations of the Simplex method have been shown to be subject to computational instabilities, which in practice often result in failure to achieve a solution to a basically well-determined problem. A numerically stable form of the Simplex method is presented with storage requirements and computational efficiency comparable with those of the standard form. The method admits non-Simplex steps and this feature enables it to be readily generalized to quadratic and nonlinear programming. Although the principal concern in this paper is not with constraints having a large number of zero elements, all necessary modification formulae are given for the extension to these cases.     

Solving a bi-objective nurse rerostering problem by using a utopic Pareto genetic heuristic

Nurse rerostering arises when at least one nurse announces that she will be unable to undertake the tasks previously assigned to her. The problem amounts to building a new roster that satisfies the hard constraints already met by the current one and, as much as possible, fulfils two groups of soft constraints which define the two objectives to be attained. A bi-objective genetic heuristic was designed on the basis of a population of individuals characterised by pairs of chromosomes, whose fitness complies with the Pareto ranking of the respective decoded solution. It includes an elitist policy, as well as a new utopic strategy, introduced for purposes of diversification. The computational experiments produced promising results for the practical application of this approach to real life instances arising from a public hospital in Lisbon.     

On solving stochastic production planning problems via scenario modelling

Summary Linear Porgramming models for stochastic planning problems and a methodology for solving them are proposed. A production planning problem with uncertainty in demand is used as a test case, but the methodology presented here is applicable to other types of problems as well. In these models, uncertainty in demand is characterized via scenarios. Solutions are obtained for each scenario and then these individual scenario solutions are aggregated to yield an implementable non-anticipative policy. Such an approach makes it possible to model correlated and nonstationary demand as well as a variety of recourse decision types. For computational purposes, two alternative representations are proposed. A compact approach that is suitable for the Simplex method and a splitting variable approach that is suitable for the Interior Point Methods. A crash procedure that generates an advanced starting solution for the Simplex method is developed. Computational results are reported with both the representations. Although some of the models presented here are very large (over 25000 constraints and 75000 variables), our computational experience with these problems is quite encouraging.     

The Complex Interior-Boundary method for linear and nonlinear programming with linear constraints

In this paper we develop the Complex method; an algorithm for solving linear programming (LP) problems with interior search directions. The Complex Interior-Boundary method (as the name suggests) moves in the interior of the feasible region from one boundary point to another of the feasible region bypassing several extreme points at a time. These directions of movement are guaranteed to improve the objective function. As a result, the Complex method aims to reach the optimal point faster than the Simplex method on large LP programs. The method also extends to nonlinear programming (NLP) with linear constraints as compared to the generalized-reduced gradient.The Complex method is based on a pivoting operation which is computationally efficient operation compared to some interior-point methods. In addition, our algorithm offers more flexibility in choosing the search direction than other pivoting methods (such as reduced gradient methods). The interior direction of movement aims at reducing the number of iterations and running time to obtain the optimal solution of the LP problem compared to the Simplex method. Furthermore, this method is advantageous to Simplex and other convex programs in regard to starting at a Basic Feasible Solution (BFS); i.e. the method has the ability to start at any given feasible solution.Preliminary testing shows that the reduction in the computational effort is promising compared to the Simplex method.     

A direct solution framework for structural optimization problems with interval uncertainties

A direct solution framework based on multi-objective evolutionary algorithm is developed to solve the structural optimization problems with interval uncertainties. The midpoint and radius of the uncertain original objective are treated as two equally important objectives, which are solved by a multi-objective evolutionary algorithm. The satisfaction value of interval possibility degree model is utilized to deal with nonlinear uncertain constraints and then the degree of constraint violation based on this model is calculated to judge the design vector individuals which one is feasible or infeasible. Subsequently, a selection strategy based on interval constrained-domination rule is utilized to realize the ranking of different design vectors. Finally, two numerical examples and the structural design of augmented reality glasses are investigated to verify the applicability and effectiveness of the proposed method.     

Implied constraints and an alternate unified development of nonlinear programming theory

This paper offers an alternate unified view of nonlinear programming theory from the perspective of implied constraints. Optimality is identically characterized for both constrained and unconstrained problems in terms of implied constraints. It is shown that there is a weaker condition than the Guignard constraint qualification for the existence of finite multipliers in the Karush-Kuhn-Tucker conditions. Surprisingly, this condition does not directly qualify the constraints but instead qualifies the objective in terms of implied constraints. More surprisingly, the existence of the finite multipliers follows directly from this objective qualification — it is not necessary for the point to be a local optimum. Methods for generating implied constraints are used to obtain a more general sufficient condition for local and global optimality. A single unified formulation of duality shows that duality is nothing more than an effort to generate the tightest implied constraint. Duality theorems hold in general for this formulation — convexity is not required — and the existence of the duality gap in prior formulations is easily explained. The algorithmic potential of this approach is highlighted by showing that the Simplex method systematically tries to imply the objective from the constraints of the problem.     

A Weighted Simplex Procedure for the Solution of Simultaneous Non-Linear Equations

The Weighted Simplex procedure, a version of the secant algorithm,provides an alternative to the Newton-Raphson procedure in thesolution of a set of non-linear equations. The results of trialsshow that, given a sufficiently good initial approximation toa solution, the two procedures converge with comparable efficiencies.Unlike the Newton-Raphson procedure the Weighted Simplex proceduredoes not involve the computation of partial derivatives.     

Dual multilevel optimization

We study the structure of dual optimization problems associated with linear constraints, bounds on the variables, and separable cost. We show how the separability of the dual cost function is related to the sparsity structure of the linear equations. As a result, techniques for ordering sparse matrices based on nested dissection or graph partitioning can be used to decompose a dual optimization problem into independent subproblems that could be solved in parallel. The performance of a multilevel implementation of the Dual Active Set algorithm is compared with CPLEX Simplex and Barrier codes using Netlib linear programming test problems.      

A simple direct cosine simplex algorithm

Linear programming (LP) is the core model of constrained optimization. The Simplex method (Simplex in short) has been proven in practice to perform very well in small- or medium-sized LP problems. A new algorithm called the direct cosine Simplex algorithm (DCA) is presented here to improve upon Simplex and to solve LP problems. The proposed DCA implements a specific cosine criterion to choose the entering variable instead of the traditional most negative rule used in Simplex. Three examples are given to illustrate the implementation of the proposed DCA to improve Simplex and to serve as the optimization tool. The utility of the proposed approach is evident from the extensive computational results on test problems adapted from NETLIB. DCA reduced the number of iterations of Simplex in most cases in our computational experiment. Preliminary results for medium-sized problems are encouraging.     

Model-based control strategies for systems with constraints of the program type

The paper presents a model-based tracking control strategy for constrained mechanical systems. Constraints we consider can be material and non-material ones referred to as program constraints. The program constraint equations represent tasks put upon system motions and they can be differential equations of orders higher than one or two, and be non-integrable. The tracking control strategy relies upon two dynamic models: a reference model, which is a dynamic model of a system with arbitrary order differential constraints and a dynamic control model. The reference model serves as a motion planner, which generates inputs to the dynamic control model. It is based upon a generalized program motion equations (GPME) method. The method enables to combine material and program constraints and merge them both into the motion equations. Lagrange’s equations with multipliers are the peculiar case of the GPME, since they can be applied to systems with constraints of first orders. Our tracking strategy referred to as a model reference program motion tracking control strategy enables tracking of any program motion predefined by the program constraints. It extends the “trajectory tracking” to the “program motion tracking”. We also demonstrate that our tracking strategy can be extended to a hybrid program motion/force tracking.     

Degenerate Linear Multicriteria Sensitivity

Sensitivity to perturbations of the vector objective function and/or the constraints is studied for the efficient (or weakly) set associated to a linear multicriteria program whose the feasible polyhedron is not necessarily assumed to be nondegenerate. Namely, the efficient set in the linear case being a connected union of some faces of the polyhedron, then we establish firstly second order necessary and sufficient conditions for the feasibility of a degenerate vertex under small perturbations, and finally, we give necessary and sufficient conditions of Simplex type, for a such vertex, or incident face to it, to be efficient and mostly to preserve this quality after the perturbations.     

Analytic efficient solution set for multi-criteria quadratic programs

We present active set methods to evaluate the exact analytic efficient solution set for multi-criteria convex quadratic programming problems (MCQP) subject to linear constraints. The idea is based on the observations that a strictly convex programming problem admits a unique solution, and that the efficient solution set for a multi-criteria strictly convex quadratic programming problem with linear equality constraints can be parameterized. The case of bi-criteria quadratic programming (BCQP) is first discussed since many of the underlying ideas can be explained much more clearly in the case of two objectives. In particular we note that the efficient solution set of a BCQP problem is a curve on the surface of a polytope. The extension to problems with more than two objectives is straightforward albeit some slightly more complicated notation. Two numerical examples are given to illustrate the proposed methods.     

Fuzzy bilevel programming with multiple objectives and cooperative multiple followers

Classic bilevel programming deals with two level hierarchical optimization problems in which the leader attempts to optimize his/her objective, subject to a set of constraints and his/her follower’s solution. In modelling a real-world bilevel decision problem, some uncertain coefficients often appear in the objective functions and/or constraints of the leader and/or the follower. Also, the leader and the follower may have multiple conflicting objectives that should be optimized simultaneously. Furthermore, multiple followers may be involved in a decision problem and work cooperatively according to each of the possible decisions made by the leader, but with different objectives and/or constraints. Following our previous work, this study proposes a set of models to describe such fuzzy multi-objective, multi-follower (cooperative) bilevel programming problems. We then develop an approximation Kth-best algorithm to solve the problems.     

A Linear Approximation for Chance-Constrained Programming

Decision environments involve the need to solve problems with varying degrees of uncertainty as well as multiple, potentially conflicting objectives. Chance constraints consider the uncertainty encountered. Codes incorporating chance constraints into a mathematical programming model are not available on a widespread basis owing to the non-linear form of the chance constraints. Therefore, accurate linear approximations would be useful to analyse this class of problems with efficient linear codes. This paper presents an approximation formula for chance constraints which can be used in either the single- or multiple-objective case. The approximation presented will place a bound on the chance constraint at least as tight as the true non-linear form, thus overachieving the chance constraint at the expense of other constraints or objectives.     

Time-restricted batching models and solution approaches for integrated forward and return product flow handling in warehouses

With the rise of Internet sales, retailers witness increases in returned products, generally because website pictures and specifications tend to be insufficient for customers to determine the right size, colour or suitability of a product. Efficient processing of the return flow can improve inventory management and increase the utilisation of warehouse capacities. Moreover, short response times to customer orders and on-time delivery are critical and shape warehouse operations. To address the problem of integrated order and return job batching, this article proposes two optimisation models with constraints on the delivery time. We propose different objectives in offline and online contexts which require unique solution approaches. Time constraints also need to be taken into account to exploit the different characteristics of order and return jobs, as well as to incorporate response time-oriented objectives. Computational examples confirm the effectiveness of included returns and the impact of consumer-oriented objectives. Finally, a comparison of the proposed solution approaches with batching procedures used by a library warehouse indicates significant savings in travel distance through integrated order and return job processing.     

A regularized decomposition method for minimizing a sum of polyhedral functions

A problem of minimizing a sum of many convex piecewise-linear functions is considered. In view of applications to two-stage linear programming, where objectives are marginal values of lower level problems, it is assumed that domains of objectives may be proper polyhedral subsets of the space of decision variables and are defined by piecewise-linear induced feasibility constraints. We propose a new decomposition method that may start from an arbitrary point and simultaneously processes objective and feasibility cuts for each component. The master program is augmented with a quadratic regularizing term and comprises an a priori bounded number of cuts. The method goes through nonbasic points, in general, and is finitely convergent without any nondegeneracy assumptions. Next, we present a special technique for solving the regularized master problem that uses an active set strategy and QR factorization and exploits the structure of the master. Finally, some numerical evidence is given.On leave from Instytut Automatyki, Politechnika Warszawska, Poland.     

Lagrange multiplier rules for non-differentiable DC generalized semi-infinite programming problems

This paper aims to study a broad class of generalized semi-infinite programming problems with (upper and lower level) objectives given as the difference of two convex functions, and (lower level) constraints described by a finite number of convex inequalities and a set constraints. First, we are interested in some various lower level constraint qualifications for the problem. Then, the results are used to establish efficient upper estimate of certain subdifferential of value functions. Finally, we apply the obtained subdifferential estimates to derive necessary optimality conditions for the problem.