Operations Research Letters 33 (2005) 475–480 Operations Research Letters www.elsevier.com/locate/orl Compoundedgeneticalgorithmsforthequadratic assignmentproblem Zvi Drezner ∗ Department of ISDS, College of Business and Economics, California State University-Fullerton, Fullerton, CA 92834-6848, USA Received 24 February 2004; accepted 21 October 2004 Available online 16 December 2004 Abstract We introduce the compounded genetic algorithm. We propose to run a quick genetic algorithm several times as Phase 1, and compile the best solutions in each run to create a starting population for Phase 2. This new approach was tested on the quadratic assignment problem with very good results. © 2004 Elsevier B.V. All rights reserved. Keywords: Quadratic assignment; Genetic algorithm; Compounded approach 1. Introduction In this paper we introduce the compounded ap- proach for genetic algorithms and test it on the quadratic assignment problem (QAP). The hybrid ge- netic algorithm (HGA) applied in the computational tests is based on the approach suggested in [5]. A few improvements to HGA are also suggested here. The quadratic assignment problem is considered to be one of the most difficult optimization problems to solve optimally. The problem is defined as follows. A set of n possible sites are given and n facilities are to be located on these sites, one facility at a site. Let c ij be the cost per unit distance between facilities i and j and d ij be the distance between sites i and j. The ∗ Tel.: +17142782712; fax: +17142785940. E-mail address: zdrezner@fullerton.edu (Z. Drezner). 0167-6377/$-see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.orl.2004.11.001 cost f to be minimized over all possible permutations, calculated for an assignment of facility i to site p(i) for i = 1,...,n, is f = n  i =1 n  j =1 c ij d p(i)p(j) . (1) For a literature review of the QAP the reader is referred to [2,3,5,7,8]. The main contribution of this paper compared with [4–6] is the novel concept of the compounded ge- netic algorithm. We also fine tuned the parameters of the hybrid genetic algorithm by suggesting flexible number of levels in the concentric tabu search and the short search which applies shorter depth in the radial search. The tandem concentric tabu search was introduced in [4]. The performance of the algorithms proposed in this paper are better than those reported in [1,4–6].