Combining metaheuristics with mathematical programming, constraint programming and machine learning

During the last years, interest on hybrid metaheuristics has risen considerably in the field of optimization and machine learning. The best results found for many optimization problems in science and industry are obtained by hybrid optimization algorithms. Combinations of optimization tools such as metaheuristics, mathematical programming, constraint programming and machine learning, have provided very efficient optimization algorithms. Four different types of combinations are considered in this paper: (i) Combining metaheuristics with complementary metaheuristics. (ii) Combining metaheuristics with exact methods from mathematical programming approaches which are mostly used in the operations research community. (iii) Combining metaheuristics with constraint programming approaches developed in the artificial intelligence community. (iv) Combining metaheuristics with machine learning and data mining techniques.     

Combining metaheuristics with mathematical programming,constraint programming and machine learning

We consider the problem of creating fair course timetables in the setting of a university. The central idea is that undesirable arrangements in the course timetable, i.e., violations of soft constraints, should be distributed in a fair way among the stakeholders. We propose and discuss in detail two fair versions of the popular curriculum-based course timetabling (CB-CTT) problem, the MMF-CB-CTT problem and the JFI-CB-CTT problem, which are based on max–min fairness (MMF) and Jain’s fairness index (JFI), respectively. For solving the MMF-CB-CTT problem, we present and experimentally evaluate an optimization algorithm based on simulated annealing. We introduce three different energy difference measures and evaluate their impact on the overall algorithm performance. The proposed algorithm improves the fairness on 20 out of 32 standard instances compared to the known best timetables. The JFI-CB-CTT problem formulation focuses on the trade-off between fairness and the aggregated soft constraint violations. Here, our experimental evaluation shows that the known best solutions to 32 CB-CTT standard instances are quite fair with respect to JFI. Our experiments show that the fairness can often be improved at the cost of only a small increase in the overall amount of penalty.     

Stability,prediction-correction,and dynamic programming

It is proven here that a bounded perturbation of the discrete dynamic programming functional equation arising from the Bolza problem yields a bounded change in its solution. This stability property encourages the development of approximation techniques for solving such equations. One such technique, involving the backward solution of an approximate functional equation as a prediction step, followed by a forward reconstruction using true equations as a correction step, is then discussed. Bounds for the errors arising from such an approximation procedure are derived. Successive approximations is suggested, in conclusion, as a means for obtaining improved solutions.     

Algorithmic equivalence in quadratic programming,I: A least-distance programming problem

It is demonstrated that Wolfe's algorithm for finding the point of smallest Euclidean norm in a given convex polytope generates the same sequence of feasible points as does the van de Panne-Whinstonsymmetric algorithm applied to the associated quadratic programming problem. Furthermore, it is shown how the latter algorithm may be simplified for application to problems of this type.This work was supported by the National Science Foundation, Grant No. MCS-71-03341-AO4, and by the Office of Naval Research, Contract No. N00014-75-C-0267.     

Penalty functions,Newton's method,and quadratic programming

In this paper, the search directions computed by two versions of the sequential quadratic programming (SQP) algorithm are compared with that computed by attempting to minimize a quadratic penalty function by Newton's method, and it is shown that the differences are attributable to ignoring certain terms in the equation for the Newton correction. Since the effect of ignoring these terms may be to make the resultant direction a poor descent direction for the quadratic penalty function, it is argued that the latter is an inappropriate merit function for use with SQP. A method is then suggested by which these terms may be included without losing the benefits gained from the use of the orthogonal transformations derived from the constraints Jacobian.The authors wish to thank A. R. Conn and N. I. M. Gould for spirited discussions which took place when the second author spent some time at Waterloo, Ontario, Canada; they also thank L. C. W. Dixon for the clarifications that he suggested to the penultimate draft of this paper.     

Optimal stopping,exponential utility,and linear programming

This paper uses linear programming to compute an optimal policy for a stopping problem whose utility function is exponential. This is done by transforming the problem into an equivalent one having additive utility and nonnegative (not necessarily substochastic) transition matrices.Research was supported by NSF Grant ENG 76-15599.     

Lifting, tilting and fractional programming revisited

Lifting, tilting and fractional programming, though seemingly different, reduce to a common optimization problem. This connection allows us to revisit key properties of these three problems on mixed integer linear sets. We introduce a simple common framework for these problems, and extend known results from each to the other two.     

Blocks of coordinates,stochastic programming,and markets

Computational Management Science - Considered here are extremal convolutions concerned with allocative efficiency, risk sharing, or market equilibrium. Each additive term is upper semicontinuous,...     

Goal programming,compromise programming and reference point method formulations: linkages and utility interpretations

This paper looks at connections between the multi-criteria techniques of goal programming, compromise programming, and the reference point method. The utility function structure of each method is examined and interpretation of the techniques in terms of pareto efficiency and the equilibrated nature of the solution is given. Means of ensuring pareto efficiency whilst maintaining as much as possible the equilibrated character are given for a certain class of model. Methods for ensuring conceptual correctness and computational accuracy by reducing reference point models to equivalent goal programmes are given. The findings are illustrated by means of an example.     

Semi-infinite programming,duality, discretization and optimality conditions†

The aim of this article is to give a survey of some basic theory of semi-infinite programming. In particular, we discuss various approaches to derivations of duality, discretization, and first- and second-order optimality conditions. Some of the surveyed results are well known while others seem to be less noticed in that area of research.     

Multilevel programming,rational reaction sets,and efficient solutions

The prime concern in this paper is the interface between efficient solutions and rational reaction sets in multilevel programming. It is shown that there need be no rational reaction solution which is also efficient, although in the bilevel case such a solution always exists, and in the generalK-level case such a solution exists if certain conditions hold. Some theoretical properties of an extension of known methods for generating efficient solutions are given, which are important algorithmically.This work was carried out under the auspices of the University of Manchester, Manchester, England and the University of Virginia, Charlottesville, Virginia, USA. The author thanks the referees for their helpful comments on the original version.     

Lattice-free sets,multi-branch split disjunctions,and mixed-integer programming

In this paper we study the relationship between valid inequalities for mixed-integer sets, lattice-free sets associated with these inequalities and the multi-branch split cuts introduced by Li and Richard (Discret Optim 5:724–734, 2008). By analyzing $n$ -dimensional lattice-free sets, we prove that for every integer $n$ there exists a positive integer $t$ such that every facet-defining inequality of the convex hull of a mixed-integer polyhedral set with $n$ integer variables is a $t$ -branch split cut. We use this result to give a finite cutting-plane algorithm to solve mixed-integer programs. We also show that the minimum value $t$ , for which all facets of polyhedral mixed-integer sets with $n$ integer variables can be generated as $t$ -branch split cuts, grows exponentially with $n$ . In particular, when $n=3$ , we observe that not all facet-defining inequalities are 6-branch split cuts.     

Generalized dynamic programming methods in integer programming

When regarded as a shortest route problem, an integer program can be seen to have a particularly simple structure. This allows the development of an algorithm for finding thek ^th best solution to an integer programming problem with max{O(kmn), O(k logk)} operations. Apart from its value in the parametric study of an optimal solution, the approach leads to a general integer programming algorithm consisting of (1) problem relaxation, (2) solution of the relaxed problem parametrically by dynamic programming, and (3) generation ofk ^th best solutions until a feasible solution is found. Elementary methods based on duality for reducingk for a given problem relaxation are then outlined, and some examples and computational aspects are discussed.