Dragan Urošević

Variable neighborhood search and local branching

In this paper we develop a variable neighborhood search (VNS) heuristic for solving mixed-integer programs (MIPs). It uses CPLEX, the general-purpose MIP solver, as a black-box. Neighborhoods around the incumbent solution are defined by adding constraints to the original problem, as suggested in the recent local branching (LB) method of Fischetti and Lodi (Mathematical Programming Series B 2003;98:23-47). Both LB and VNS use the same tools: CPLEX and the same definition of the neighborhoods around the incumbent. However, our VNS is simpler and more systematic in neighborhood exploration. Consequently, within the same time limit, we were able to improve 14 times the best known solution from the set of 29 hard problem instances used to test LB.

A general variable neighborhood search for solving the uncapacitated single allocation p-hub median problem

We present a new general variable neighborhood search approach for the uncapacitated single allocation p-hub median problem in networks. This NP hard problem is concerned with locating hub facilities in order to minimize the traffic between all origin-destination pairs. We use three neighborhoods and efficiently update data structures for calculating new total flow in the network. In addition to the usual sequential strategy, a new nested strategy is proposed in designing a deterministic variable neighborhood descent local search. Our experimentation shows that general variable neighborhood search based heuristics outperform the best-known heuristics in terms of solution quality and computational effort. Moreover, we improve the best-known objective values for some large Australia Post and PlanetLab instances. Results with the new nested variable neighborhood descent show the best performance in solving very large test instances.

Variable neighborhood search for the maximum clique

Maximum clique is one of the most studied NP-hard optimization problem on graphs because of its simplicity and its numerous applications. A basic variable neighborhood search heuristic for maximum clique that combines greedy with the simplicial vertex test in its descent step is proposed and tested on standard test problems from the literature. Despite its simplicity, the proposed heuristic outperforms most of the well-known approximate solution methods. Moreover, it gives solution of equal quality to those of the state-of-the-art heuristic of Battiti and Protasi in half the time.

Reformulation descent applied to circle packing problems

Several years ago classical Euclidean geometry problems of densest packing of circles in the plane have been formulated as nonconvex optimization problems, allowing to find heuristic solutions by using any available NLP solver. In this paper we try to improve this procedure. The faster NLP solvers use first order information only, so stop in a stationary point. A simple switch from Cartesian coordinates to polar or vice versa, may destroy this stationarity and allow the solver to descend further. Such formulation switches may of course be iterated. For densest packing of equal circles into a unit circle, this simple feature turns out to yield results close to the best known, while beating second order methods by a time-factor well over 100.This technique is formalized as a general reformulation descent (RD) heuristic, which iterates among several formulations of the same problem until local searches obtain no further improvement. We also briefly discuss how RD might be used within other metaheuristic schemes.

A general variable neighborhood search for the one-commodity pickup-and-delivery travelling salesman problem

We present a variable neighborhood search approach for solving the one-commodity pickup-and-delivery travelling salesman problem. It is characterized by a set of customers such that each of the customers either supplies (pickup customers) or demands (delivery customers) a given amount of a single product, and by a vehicle, whose given capacity must not be exceeded, that starts at the depot and must visit each customer only once. The objective is to minimize the total length of the tour. Thus, the considered problem includes checking the existence of a feasible travelling salesman’s tour and designing the optimal travelling salesman’s tour, which are both NP-hard problems. We adapt a collection of neighborhood structures, k-opt, double-bridge and insertion operators mainly used for solving the classical travelling salesman problem. A binary indexed tree data structure is used, which enables efficient feasibility checking and updating of solutions in these neighborhoods. Our extensive computational analysis shows that the proposed variable neighborhood search based heuristics outperforms the best-known algorithms in terms of both the solution quality and computational efforts. Moreover, we improve the best-known solutions of all benchmark instances from the literature (with 200 to 500 customers). We are also able to solve instances with up to 1000 customers.

Variable neighbourhood decomposition search for 0-1 mixed integer programs

In this paper we propose a new hybrid heuristic for solving 0-1 mixed integer programs based on the principle of variable neighbourhood decomposition search. It combines variable neighbourhood search with a general-purpose CPLEX MIP solver. We perform systematic hard variable fixing (or diving) following the variable neighbourhood search rules. The variables to be fixed are chosen according to their distance from the corresponding linear relaxation solution values. If there is an improvement, variable neighbourhood descent branching is performed as the local search in the whole solution space. Numerical experiments have proven that exploiting boundary effects in this way considerably improves solution quality. With our approach, we have managed to improve the best known published results for 8 out of 29 instances from a well-known class of very difficult MIP problems. Moreover, computational results show that our method outperforms the CPLEX MIP solver, as well as three other recent most successful MIP solution methods.

Primal-Dual Variable Neighborhood Search for the Simple Plant-Location Problem

The variable neighborhood search metaheuristic is applied to the primal simple plant-location problem and to a reduced dual obtained by exploiting the complementary slackness conditions. This leads to (i) heuristic resolution of (metric) instances with uniform fixed costs, up to n=15,000 users, and m=n potential locations for facilities with an error not exceeding 0.04%; (ii) exact solution of such instances with up to m=n=7,000; and (iii) exact solutions of instances with variable fixed costs and up to m=n=15,000.

Solving large p-median clustering problems by primal–dual variable neighborhood search

Data clustering methods are used extensively in the data mining literature to detect important patterns in large datasets in the form of densely populated regions in a multi-dimensional Euclidean space. Due to the complexity of the problem and the size of the dataset, obtaining quality solutions within reasonable CPU time and memory requirements becomes the central challenge. In this paper, we solve the clustering problem as a large scale p-median model, using a new approach based on the variable neighborhood search (VNS) metaheuristic. Using a highly efficient data structure and local updating procedure taken from the OR literature, our VNS procedure is able to tackle large datasets directly without the need for data reduction or sampling as employed in certain popular methods. Computational results demonstrate that our VNS heuristic outperforms other local search based methods such as CLARA and CLARANS even after upgrading these procedures with the same efficient data structures and local search. We also obtain a bound on the quality of the solutions by solving heuristically a dual relaxation of the problem, thus introducing an important capability to the solution process.

Variable neighbourhood search for bandwidth reduction

The problem of reducing the bandwidth of a matrix consists of finding a permutation of rows and columns of a given matrix which keeps the non-zero elements in a band as close as possible to the main diagonal. This NP-complete problem can also be formulated as a vertex labelling problem on a graph, where each edge represents a non-zero element of the matrix. We propose a variable neighbourhood search based heuristic for reducing the bandwidth of a matrix which successfully combines several recent ideas from the literature. Empirical results for an often used collection of 113 benchmark instances indicate that the proposed heuristic compares favourably to all previous methods. Moreover, with our approach, we improve best solutions in 50% of instances of large benchmark tests.

Variable neighborhood decomposition search for the edge weighted k -cardinality tree problem

The minimum k-cardinality tree problem on graph G consists in finding a subtree of G with exactly k edges whose sum of weights is minimum. A number of heuristic methods have been developed recently to solve this NP-hard problem. In this paper a decomposition approach is developed and implemented within a successive approximation scheme known as variable neighborhood decomposition search. This approach obtains superior results over existing methods, and furthermore, allows larger problem instances(up to 5000 nodes) to be solved more efficiently.