## The Quadratic Assignment Problem: A Survey and Recent Developments (1994)

Venue: | In Proceedings of the DIMACS Workshop on Quadratic Assignment Problems, volume 16 of DIMACS Series in Discrete Mathematics and Theoretical Computer Science |

Citations: | 90 - 16 self |

### BibTeX

@INPROCEEDINGS{Pardalos94thequadratic,

author = {Panos M. Pardalos and Franz Rendl and Henry Wolkowicz},

title = {The Quadratic Assignment Problem: A Survey and Recent Developments},

booktitle = {In Proceedings of the DIMACS Workshop on Quadratic Assignment Problems, volume 16 of DIMACS Series in Discrete Mathematics and Theoretical Computer Science},

year = {1994},

pages = {1--42},

publisher = {American Mathematical Society}

}

### Years of Citing Articles

### OpenURL

### Abstract

. Quadratic Assignment Problems model many applications in diverse areas such as operations research, parallel and distributed computing, and combinatorial data analysis. In this paper we survey some of the most important techniques, applications, and methods regarding the quadratic assignment problem. We focus our attention on recent developments. 1. Introduction Given a set N = f1; 2; : : : ; ng and n \Theta n matrices F = (f ij ) and D = (d kl ), the quadratic assignment problem (QAP) can be stated as follows: min p2\Pi N n X i=1 n X j=1 f ij d p(i)p(j) + n X i=1 c ip(i) ; where \Pi N is the set of all permutations of N . One of the major applications of the QAP is in location theory where the matrix F = (f ij ) is the flow matrix, i.e. f ij is the flow of materials from facility i to facility j, and D = (d kl ) is the distance matrix, i.e. d kl represents the distance from location k to location l [62, 67, 137]. The cost of simultaneously assigning facility i to locat...

### Citations

10921 | Computers and Intractability: A Guide to the Theory of NP-Completeness - Garey, Johnson |

4675 | Matrix Analysis - HORN, JOHNSON - 1990 |

1459 | Theory of linear and integer programming - Schrijver - 1986 |

1430 |
Reducibility Among Combinatorial Problems
- KARP
- 1972
(Show Context)
Citation Context ...t the number of variables of the three-dimensional assignment problem of order n is n 3 . From the complexity point of view, it has been shown that the three-dimensional assignment problem is NP-hard =-=[131]-=-. Most ot the proposed algorithms for this problem are implicit enumeration methods. Some of the proposed algorithms include those of Vlach [247], Pierskalla [196, 197] and Leue [145]; a primaldual al... |

1271 | Combinatorial Optimization: Algorithms and Complexity - Papadimitriou, Steiglitz - 1982 |

1139 | Geometric Algorithms and Combinatorial Optimizations - Grotschel, Lovász, et al. - 1993 |

1046 | An efficient heuristic procedure for partitioning graphs - Kernighan, Lin - 1970 |

1038 | Integer and Combinatorial Optimization - Nemhauser, Wolsey - 1988 |

510 | Optimization by simulated annealing: an experimental evaluation - Johnson, Aragon, et al. - 1989 |

480 | Greedy randomized adaptive search procedures
- Feo, Resende
- 1995
(Show Context)
Citation Context ... via an adaptive randomized greedy function; the second applies a local search technique to the constructed solution in hope of finding an improvement. A comprehensive survey of GRASP can be found in =-=[73]-=-. In [71] GRASP has been applied to a quadratic assignment problem that models the positioning of intermodal highway trailers on railcars. The GRASP is THE QUADRATIC ASSIGNMENT PROBLEM 27 incorporated... |

322 | An effective heuristic algorithm for the traveling salesman problem - Lin, Kernighan - 1973 |

291 |
Circulant Matrices
- Davis
- 1979
(Show Context)
Citation Context ...unds, Section (5.2.2).) In addition, one can parametrize the orthogonal constraint X 2 O by noting that it is equivalent to the matrix exponential X = exp S for some skew symmetric matrix S, see e.g. =-=[172, 60]. Th-=-is results in an unconstrained problem over the space of skew symmetric matrices. 2.3.5. Semidefinite Programming. The orthogonal constraint can be relaxed to XX t �� I, i.e. XX t is negative semi... |

278 |
P-complete approximation problems
- Sahni, Gonzales
- 1976
(Show Context)
Citation Context ...ems and their solutions. 4.1. Computational Complexity. In 1976, Sahni and Gonzalez showed that the QAP is NP -complete, which implies that finding a polynomial-time algorithm to solve it is unlikely =-=[221]-=-. In addition, they have also shown that QAP belongs even to the hardest core of this complexity class, in the sense that the problem of finding an ffl-approximate solution of QAP remains NP -complete... |

261 | The Travelling Salesman Problem: A Guided Tour of Combinatorial Optimization - Lawler, Lenstra, et al. - 1990 |

237 |
Kronecker Products and Matrix Calculus: with Applications
- Graham
- 1981
(Show Context)
Citation Context ... in that we do not take advantage of the hidden fact that we are really working on an n dimensional problem. For that reason, the Kronecker product is rarely used and we do not study it further. (See =-=[96]-=- for details on manipulations and calculus involving Kronecker products.) 2.2. The Feasible Set and Perturbations. The feasible set for QAP consists of all the possible assignments of n objects to n l... |

217 | The zero-error capacity of a noisy channel - Shannon - 1956 |

196 |
How easy is local search
- Johnson, Papadimitriou, et al.
- 1988
(Show Context)
Citation Context ...nt is possible. To characterize the complexity of solving combinatorial optimization problems such as the QAP with local search algorithms, a Polynomial-time Local Search (PLS) class has been defined =-=[126]-=- that captures the structure of NP problems at the level of their feasible solutions and neighborhoods. Similar to NP-completeness, the concept of PLS-completeness has been defined to capture the clas... |

162 | QAPLIB-A Quadratic assignment problem library”, http://www.opt.math.tu-graz.ac.at/qaplib. www.ijacsa.thesai.org 5 | P a g e
- Bedkard, Karisch, et al.
- 2014
(Show Context)
Citation Context ...body of the message put "send 92006"). 6.3. QAP-LIB. Finally we point out that a collection of more than 130 instances of quadratic assignment problems is contained in a library called the &=-=quot;QAP-LIB", [39]-=-. This library consists of two parts. The data part contains various instances given by the input matrices A, B and C, if C 6= 0. Then THE QUADRATIC ASSIGNMENT PROBLEM 31 there is a documentation, cor... |

157 |
19810n the mapping problem
- Bokharï
(Show Context)
Citation Context ...Cases. The quadratic assignment problem can be formulated very naturally in a graph theoretical context. This formulation was investigated first by Christofides and Gerrard [54], and later by Bokhari =-=[23]-=- and Rendl [208]. We review this formulation and present several applications to other optimization problems on graphs that can be derived from this formulation. Let G = (V; E) and G 0 = (V 0 ; E 0 ) ... |

146 | A probabilistic heuristic for a computationally difficult set covering problem - Feo, Resende |

128 |
Parallel genetic algorithms, population genetics and combinatorial optimisation
- Muhlenbein
- 1989
(Show Context)
Citation Context ...ly parallel. A number of researchers have tried to apply genetic algorithms to solve combinatorial optimization problems, such as the the graph partitioning problem and the traveling salesman problem =-=[161]-=-. 5.4. Greedy Randomized Adaptive Search Procedures (GRASP). GRASP is an iterative randomized sampling technique in which each iteration provides an approximate solution to the problem at hand. The in... |

123 |
Combinatorial Matrix Theory
- Brualdi, Ryser
- 1991
(Show Context)
Citation Context ...ces is the set of doubly stochastic matrices, conv \Pi = D: Thus the set of doubly stochastic matrices corresponds to the bipartite perfect matching polytope. (For more details on the above, see e.g. =-=[28]-=-.) The properties of the feasible set allow for perturbations of the objective function without changing the optimal solution of the original problem. These perturbations do change relaxations for the... |

121 |
Pardalos. A collection of test problems for constrained global optimization algorithms
- Floudas, M
- 1990
(Show Context)
Citation Context ... platform on which different algorithms for the same problem can be compared. For general references on this subject, see e.g., Pardalos [178, 179], Pardalos and Rosen [190], and Floudas and Pardalos =-=[76]-=-. 6.1. Palubetskis' Generator for QAPs with a Known Solution. Next, we discuss the generation of test problems for the quadratic assignment problem. One of the first methods for constructing test prob... |

116 | Graph Theory with Applications to Engineering and Computer - Deo - 2000 |

101 | The maximum clique problem - Pardalos, Xue - 1994 |

100 | On the facial structure of set packing polyhedra - Padberg - 1973 |

86 | O.: Evolution algorithms in combinatorial optimization - Muhlenbein, Georges-Schleuter, et al. - 1988 |

82 |
Simple local search problems that are hard to solve
- er, Yannakakis
- 1991
(Show Context)
Citation Context ...t of PLS-completeness has been defined to capture the class of the hardest problems in PLS. For certain NP-complete problems, the corresponding PLS problems have already been shown to be PLS-complete =-=[126, 223]-=-. In regard to the complexity of local search, see also [181] and [191]. 4.2.1. A K-L Type Local Search Algorithm. Next we describe a new local search algorithm for the QAP and establish the connectio... |

79 |
An experimental comparison of techniques for the assignment of facilities to locations
- Nugent, Vollmann, et al.
- 1968
(Show Context)
Citation Context ... earliest branch and bound algorithms for solving QAPs are described in [36], [66], [180] and [218]. Pair assignment algorithms were developed by Gavett and Plyter [86], Land [141], and Nugent et al. =-=[171]-=-, etc. At each node of the branch-and-bound search tree, a fixed pair of facilities is allocated to a pair of locations. The last algorithm, the relative positioning algorithm, was developed by Mircha... |

78 | Large-scale optimization of eigenvalues
- Overton
- 1992
(Show Context)
Citation Context ...unds, Section (5.2.2).) In addition, one can parametrize the orthogonal constraint X 2 O by noting that it is equivalent to the matrix exponential X = exp S for some skew symmetric matrix S, see e.g. =-=[172, 60]. Th-=-is results in an unconstrained problem over the space of skew symmetric matrices. 2.3.5. Semidefinite Programming. The orthogonal constraint can be relaxed to XX t �� I, i.e. XX t is negative semi... |

76 |
Optimal and suboptimal algorithms for the quadratic assignment problem
- GILMORE
- 1962
(Show Context)
Citation Context ...ty permutation as the initial partial permutation; during the execution of the algorithms, the partial permutation is extended to a full permutation. Single assignment algorithms date back to Gilmore =-=[89]-=- and were extended to the general QAP by Lawler [143]. In this approach, as described in the work by Gilmore and Lawler, a facility is assigned to a location at each node of the branch-and-bound searc... |

68 |
A thermodynamically motivated Simulation procedure for combinatorial optimization problems", European
- Burkard, Rendle
- 1989
(Show Context)
Citation Context ...normal steepest descent strategy. However, it is argued in the analogy that by taking such controlled ascent steps, the optimization algorithm needs not get stuck on poor solutions. Burkard and Rendl =-=[41]-=- applied simulated annealing to the QAP and reported favorable computational results. Wilhelm and Ward [253] further investigated the procedure. Essential to the success of the adaptation of simulated... |

65 | A More Portable Fortran Random Number Generator - Schrage - 1979 |

64 | Computationally related problems - Sahni - 1974 |

63 |
The backboard wiring problem: A placement algorithm
- Steinberg
- 1961
(Show Context)
Citation Context ...this location problem. In addition to its application in facility location problems, the QAP has been found useful in such applications as scheduling [88], the backboard wiring problem in electronics =-=[240], parallel-=- and distributed computing [24], and statistical data analysis [118]. Other applications may be found in [77, 138, 159]. The term "quadratic" comes from the reformulation of the problem as a... |

60 |
Robust tabu search for the quadratic assignment problem. Parallel Comput
- Taillard
(Show Context)
Citation Context ...approach has been applied successfully to a number of combinatorial optimization problems including the TSP. Adaptations of tabu search to the QAP have been studied by Skorin-Kapov [238] and Taillard =-=[242]-=-. The basic idea is as follows. To improve a given initial permutation tabu search seeks, among the set of permutations obtained by a pair exchange of assignments, a permutation with the best heuristi... |

57 | Finding a maximum clique in an arbitrary graph - Balas, Yu - 1986 |

57 | Indefinite trust region subproblems and nonsymmetric eigenvalue perturbations
- Stern, Wolkowicz
- 1995
(Show Context)
Citation Context ...n in (2.1) and relax the constraint set to X 2 S " E ; then we do not have to worry about convexity of the objective function, i.e. we obtain a tractable problem called a trust region subproblem.=-= (In [241]-=- it is shown that these problems are really implicit convex problems, since their dual problems are concave maximization problems.) However, these problems still do not exploit the structure of QAP. 2... |

55 |
The bandwidth problem for graphs and matrices - a survey
- Chinn, Chvátalová, et al.
- 1982
(Show Context)
Citation Context ..., see [75]. Perhaps less known are the connections of the quadratic assignment problem to the bandwidth problem in graphs. To see this connection we first introduce the bandwidth problem (we refer to =-=[201] for a s-=-urvey on the topic): Let G be an undirected (and unweighted) graph on n nodes. A permutation �� of n elements is called a labeling of the nodes of G. The bandwidth of a labeling �� is defined ... |

51 |
Heuristic Algorithm and Simulation Approach to Relative Location of Facilities
- Armour, Buffa
- 1963
(Show Context)
Citation Context ...) is the set fp(i) j i 2 Mg. This process is repeated until p becomes a complete permutation. One of the oldest heuristics used is the CRAFT (Computerized Relative Allocation of Facilities Technique) =-=[4, 231, 30]-=-. This is a well-known heuristic for designing the layout of facilities that has been in use for over 25 years. Given a set of departments, locations, a matrix of flows between departments, and a matr... |

50 |
Quadratic assignment problems
- Finke, Burkard, et al.
(Show Context)
Citation Context ...re are roughly 3 categories of lower bounds. The first category includes the classical GilmoreLawler bound (GLB) and related bounds [89, 143]. The second category includes the eigenvalue based bounds =-=[74, 102, 101, 210, 103]-=-. The rest of the bounds are mostly based on reformulations of the QAP and generally involves solving a number of linear assignment problems [5, 47, 54, 82]. It is generally acknowledged that the eige... |

47 | An analysis of some graph theoretical cluster techniques - Augustson, Minker - 1970 |

47 |
Tabu Search - Part 1
- Glover
- 1989
(Show Context)
Citation Context ...etter solution in the neighborhood of the current solution. The algorithm terminates when there is no better solution in the neighborhood of the current solution. Tabu search was introduced by Glover =-=[90, 91]-=- as a technique to overcome local optimality in combinatorial search. The underlying idea is to limit the search directions for each search step to obtain good quality solutions in an effective way. T... |

45 | On the ratio of optimal integral and fractional covers, Discrete Mathematics 13 - Lovasz - 1975 |

44 |
The complexity of facets resolved
- Papadimitriou, Wolfe
- 1988
(Show Context)
Citation Context ...n to the global optimum. From the complexity point of view, it can be shown that, if there exists a polynomial time algorithm for checking whether a given permutation is globally optimal, then P = NP =-=[177]-=-. 18 P.M. PARDALOS, F. RENDL, AND H. WOLKOWICZ 4.4. Asymptotic Behavior. A nice feature of the QAP is that the relative difference between the worst and optimal solutions becomes arbitrarily small wit... |

43 | The Linear Ordering Problem: Algorithms and Applications, Research and Exposition - Reinelt - 1985 |

41 | Facilities Location: Models and Methods - Love, Morris, et al. - 1988 |

41 | A Projection Technique for Partitioning the Nodes of Graph
- Rendl, Wolkowicz
- 1990
(Show Context)
Citation Context ...ons ��. Finally the problem of determining whether a vertex separator of some given size k exists in a graph can be modeled as a quadratic assignment problem. We refer to further details in [110].=-= In [166, 209]-=- it is pointed out that the general graph partition problem can be modeled as a quadratic assignment problem. (See Section 4.3 for details.) It turns out however, that exploiting the special structure... |

40 | A simplified algorithm for obtaining approximate solutions to zero-one programming problems, Mgml Sci - Toyoda - 1975 |

39 |
Maximum properties and inequalities for the eigenvalues of completely continuous operators
- Fan
(Show Context)
Citation Context ...(2.1.2). A lower bound for the quadratic part of QAP, based on eigenvalues of A and B, is given by relaxing the constraint set of permutation matrices to O. This results in the following theorem (see =-=[70, 74]). Theorem 5.1. -=-Let A and B be symmetric matrices, ands1s2 :::sn be the eigenvalues of A, and �� 1s�� 2 :::s��n be the eigenvalues of B. For any p 2 \Pi, n X i=1si ��n\Gammai+1sn X i=1 n X j=1 a ij b ... |