Handbooks in Operations Research and Management Science
Discrete OptimizationEdited by
- K. Aardal, Centrum voor Wiskunde en Informatica, 1090 GB, Amsterdam, The Netherlands
- George Nemhauser, School of Industrial & Systems Engineering, Georgia Institute of Technology, Atlanta, GA, USA
- R. Weismantel, Otto-von-Guericke-University of Magdeburg, 39106 Magdeburg, Germany
The chapters of this Handbook volume covers nine main topics that are representative of recenttheoretical and algorithmic developments in the field. In addition to the nine papers that present the state of the art, there is an article onthe early history of the field.
The handbook will be a useful reference to experts in the field as well as students and others who want to learn about discrete optimization.
All of the chapters in this handbook are written by authors who have made significant original contributions to their topics. Herewith a brief introduction to the chapters of the handbook.
"On the history of combinatorial optimization (until 1960)" goes back to work of Monge in the 18th century on the assignment problem and presents six problem areas: assignment, transportation,maximum flow, shortest tree, shortest path and traveling salesman.The branch-and-cut algorithm of integer programming is the computational workhorse of discrete optimization. It provides the tools that have been implemented in commercial software such as CPLEXand Xpress MP that make it possible to solve practical problems in supply chain, manufacturing, telecommunications and many other areas."Computational integer programming and cutting planes" presents the key ingredientsof these algorithms.
Although branch-and-cut based on linear programming relaxation is the most widely used integer programming algorithm, other approaches areneeded to solve instances for which branch-and-cut performs poorly and to understand better the structure of integral polyhedra. The next three chapters discuss alternative approaches."The structure of group relaxations" studies a family of polyhedra obtained by dropping certainnonnegativity restrictions on integer programming problems.
Although integer programming is NP-hard in general, it is polynomially solvable in fixed dimension. "Integer programming, lattices, and results in fixed dimension" presents results in this area including algorithms that use reduced bases of integer lattices that are capable of solving certain classes of integer programs that defy solution by branch-and-cut.Relaxation or dual methods, such as cutting plane algorithms,progressively remove infeasibility while maintaining optimality to the relaxed problem. Such algorithms have the disadvantage ofpossibly obtaining feasibility only when the algorithm terminates.Primal methods for integer programs, which move from a feasible solution to a better feasible solution, were studied in the 1960'sbut did not appear to be competitive with dual methods. However,recent development in primal methods presented in "Primal integer programming" indicate that this approach is not just interesting theoretically but may have practical implications as well.
The study of matrices that yield integral polyhedra has a long tradition in integer programming. A major breakthrough occurred in the 1990's with the development of polyhedral and structural resultsand recognition algorithms for balanced matrices. "Balanced matrices" is a tutorial on thesubject.Submodular function minimization generalizes some linear combinatorial optimization problems such as minimum cut and is one of the fundamental problems of the field that is solvable in polynomialtime. "Submodular function minimization" presents the theory and algorithms of this subject.
In the search for tighter relaxations of combinatorial optimization problems, semidefinite programming provides a generalization oflinear programming that can give better approximations and is still polynomially solvable. This subject is discussed in "Semidefinite programming and integer programming".Many real world problems have uncertain data that is known only probabilistically. Stochastic programming treats this topic, but until recently it was limited, for computational reasons, tostochastic linear programs. Stochastic integer programming is now a high profile research area and recent developments are presented in"Algorithms for stochastic mixed-integer programmingmodels".Resource constrained scheduling is an example of a class of combinatorial optimization problems that is not naturally formulated with linear constraints so that linear programming based methods donot work well. "Constraint programming" presents an alternative enumerative approach that is complementary to branch-and-cut. Constraint programming,primarily designed for feasibility problems, does not use a relaxation to obtain bounds. Instead nodes of the search tree arepruned by constraint propagation, which tightens bounds on variables until their values are fixed or their domains are shown to be empty.
Handbooks in Operations Research and Management Science
Hardbound, 620 Pages
Published: December 2005
- 1. On the History of Combinatorial Optimization (till 1960) (A. Schrijver). 2. Computational Integer Programming and Cutting Planes (A. Fügenschuh, A. Martin). 3. The Structure of Group Relaxations (R. R. Thomas). 4. Integer programming, lattices, and results in fixed dimension (K. Aardal, F. Eisenbrand). 5. Primal Integer Programming (B. Spille, R. Weismantel). 6. Balanced Matrices (G. Cornuéjols, M. Conforti). 7. Submodular Function Minimization (T. McCormick). 8. Semidefinite Programming and Integer Programming (M. Laurent, F. Rendl). 9. Algorithms for Stochastic Mixed-Integer Programming Models (S. Sen). 10. Constraint Programming (A. Bockmayr, J.N. Hooker).