Solving Steiner Tree Problems in Graphs with Lagrangian Relaxation

This paper presents an algorithm to obtain near optimal solutions for the Steiner tree problem in graphs. It is based on a Lagrangian relaxation of a multi-commodity flow formulation of the problem. An extension of the subgradient algorithm, the volume algorithm, has been used to obtain lower bounds and to estimate primal solutions. It was possible to solve several difficult instances from the literature to proven optimality without branching. Computational results are reported for problems drawn from the SteinLib library.     

Generalized Steiner Problems and Other Variants

In this paper, we examine combinatorial optimization problems by considering the case where the set N (the ground set of elements) is expressed as a union of a finite number of m nonempty distinct subsets N ₁,...,N _m. The term we use is the generalized Steiner problems coined after the Generalized Traveling Salesman Problem. We have collected a short list of classical combinatorial optimization problems and we have recast each of these problems in this broader framework in an attempt to identify a linkage between these “generalized” problems. In the literature one finds generalized problems such as the Generalized Minimum Spanning Tree (GMST), Generalized Traveling Salesman Problem (GTSP) and Subset Bin-packing (SBP). Casting these problems into the new problem setting has important implications in terms of the time effort required to compute an optimal solution or a “good” solution to a problem. We examine questions like “is the GTSP “harder” than the TSP?” for a number of paradigmatic problems starting with “easy” problems such as the Minimal Spanning Tree, Assignment Problem, Chinese Postman, Two-machine Flow Shop, and followed by “hard” problems such as the Bin-packing, and the TSP.     

On Shortest k-Edge-Connected Steiner Networks in Metric Spaces

Given a set of points P in a metric space, let l(P) denote the ratio of lengths between the shortest k-edge-connected Steiner network and the shortest k-edge-connected spanning network on P, and let r = inf l(P) P for k 1. In this paper, we show that in any metric space, r 3/4 for k 2, and there exists a polynomial-time -approximation for the shortest k-edge-connected Steiner network, where = 2 for even k and = 2 + 4/(3k) for odd k. In the Euclidean plane, and .     

The Steiner Tree Problem in Kalmanson Matrices and in Circulant Matrices

We investigate the computational complexity of two special cases of the Steiner tree problem where the distance matrix is a Kalmanson matrix or a circulant matrix, respectively. For Kalmanson matrices we develop an efficient polynomial time algorithm that is based on dynamic programming. For circulant matrices we give an -hardness proof and thus establish computational intractability.     

Polynomial Time Approximation Scheme for the Rectilinear Steiner Arborescence Problem

Given a set N of n terminals in the first quadrant of the Euclidean plane E², find a minimum length directed tree rooted at the origin o, connecting to all terminals in N, and consisting of only horizontal and vertical arcs oriented from left to right or from bottom to top. This problem is called rectilinear Steiner arborescence problem, which has been proved to be NP-complete recently (Shi and Su, 11th ACM-SIAM Symposium on Discrete Algorithms (SODA), January 2000, to appear). In this paper, we present a polynomial time approximation scheme for this problem.     

Efficient Heuristics for Orientation Metric and Euclidean Steiner Tree Problems

We consider Steiner minimum trees (SMT) in the plane, where only orientations with angle , 0 i – 1 and an integer, are allowed. The orientations define a metric, called the orientation metric, , in a natural way. In particular, ₂ metric is the rectilinear metric and the Euclidean metric can beregarded as metric. In this paper, we provide a method to find an optimal SMT for 3 or 4 points by analyzing the topology of SMT's in great details. Utilizing these results and based on the idea of loop detection first proposed in Chao and Hsu, IEEE Trans. CAD, vol. 13, no. 3, pp. 303–309, 1994, we further develop an O(n²) time heuristic for the general SMT problem, including the Euclidean metric. Experiments performed on publicly available benchmark data for 12 different metrics, plus the Euclidean metric, demonstrate the efficiency of our algorithms and the quality of our results.     

On the Complexity of the Steiner Problem

Recently Rubinstein et al. gave a new proof of the NP-completeness of the discretized Steiner problem, that is, the problem of finding a shortest network connecting a given set of points in the plane where all vertices are integer points and a discretized metric is used. Their approach was to consider the complexity of the PALIMEST problem, the Steiner problem for sets of points lying on two parallel lines. In this paper, we give a new proof of this theorem, using simpler, more constructive arguments. We then extend the result to a more general class of networks known as APE-Steiner trees in which certain angles between edges or slopes of edges are specified beforehand.     

Steiner systems for two-stage disjunctive testing

The subject of this paper are some constructions of Steiner designs with blocks of two sizes that differ by one. The study of such designs is motivated by a combinatorial lower bound on the minimum number of individual tests at the second stage of a 2-stage disjunctive testing algorithm.     

Approximation Algorithms for Bounded Facility Location Problems

The bounded k-median problem is to select in an undirected graph G = (V,E) a set S of k vertices such that the distance from any vertex v V to S is at most a given bound d and the average distance from vertices V\S to S is minimized. We present randomized algorithms for several versions of this problem and we prove some inapproximability results. We also study the bounded version of the uncapacitated facility location problem and present extensions of known deterministic algorithms for the unbounded version.     

Approximation and Exact Algorithms for Constructing Minimum Ultrametric Trees from Distance Matrices

An edge-weighted tree is called ultrametric if the distances from the root to all the leaves in the tree are equal. For an n by n distance matrix M, the minimum ultrametric tree for M is an ultrametric tree T = (V, E, w) with leaf set {1,..., n} such that d_T(i, j) M[i, j] for all i, j and is minimum, where d_T(i, j) is the distance between i and j on T. Constructing minimum ultrametric trees from distance matrices is an important problem in computational biology. In this paper, we examine its computational complexity and approximability. When the distances satisfy the triangle inequality, we show that the minimum ultrametric tree problem can be approximated in polynomial time with error ratio 1.5(1 + log n), where n is the number of species. We also develop an efficient branch-and-bound algorithm for constructing the minimum ultrametric tree for both metric and non-metric inputs. The experimental results show that it can find an optimal solution for 25 species within reasonable time, while, to the best of our knowledge, there is no report of algorithms solving the problem even for 12 species.     

Two Variations of the Minimum Steiner Problem

Given a set S of starting vertices and a set T of terminating vertices in a graph G = (V,E) with non-negative weights on edges, the minimum Steiner network problem is to find a subgraph of G with the minimum total edge weight. In such a subgraph, we require that for each vertex s S and t T, there is a path from s to a terminating vertex as well as a path from a starting vertex to t. This problem can easily be proven NP-hard. For solving the minimum Steiner network problem, we first present an algorithm that runs in time and space that both are polynomial in n with constant degrees, but exponential in |S|+|T|, where n is the number of vertices in G. Then we present an algorithm that uses space that is quadratic in n and runs in time that is polynomial in n with a degree O(max {max {|S|,|T|}–2,min {|S|,|T|}–1}). In spite of this degree, we prove that the number of Steiner vertices in our solution can be as large as |S|+|T|–2. Our algorithm can enumerate all possible optimal solutions. The input graph G can either be undirected or directed acyclic. We also give a linear time algorithm for the special case when min {|S|,|T|} = 1 and max {|S|,|T|} = 2.The minimum union paths problem is similar to the minimum Steiner network problem except that we are given a set H of hitting vertices in G in addition to the sets of starting and terminating vertices. We want to find a subgraph of G with the minimum total edge weight such that the conditions required by the minimum Steiner network problem are satisfied as well as the condition that every hitting vertex is on a path from a starting vertex to a terminating vertex. Furthermore, G must be directed acyclic. For solving the minimum union paths problem, we also present algorithms that have a time and space tradeoff similar to algorithms for the minimum Steiner network problem. We also give a linear time algorithm for the special case when |S| = 1, |T| = 1 and |H| = 2.An extended abstract of part of this paper appears in Hsu et al. (1996).Supported in part by the National Science Foundation under Grants CCR-9309743 and INT-9207212, and by the Office of Naval Research under Grant No. N00014-93-1-0272.Supported in part by the National Science Council, Taiwan, ROC, under Grant No. NSC-83-0408-E-001-021.     

Finding a Noncrossing Steiner Forest in Plane Graphs Under a 2-Face Condition

Let G = (V,E) be a plane graph with nonnegative edge weights, and let be a family of k vertex sets , called nets. Then a noncrossing Steiner forest for in G is a set of k trees in G such that each tree connects all vertices, called terminals, in net N _i, any two trees in do not cross each other, and the sum of edge weights of all trees is minimum. In this paper we give an algorithm to find a noncrossing Steiner forest in a plane graph G for the case where all terminals in nets lie on any two of the face boundaries of G. The algorithm takes time if G has n vertices and each net contains a bounded number of terminals.     

Scheduling in Switching Networks with Set-Up Delays

We consider the (preemptive bipartite scheduling problem PBS) (Crescenzi et al., On approximating a scheduling problem, Journal of Combinatorial Optimization, vol. 5, pp. 287–297, 2001) arising in switching communication systems, where each input and output port can be involved in at most one communication at the same time. Given a set of communication tasks to be communicated from the transmitters to the receivers of such a system, we aim to find a schedule minimizing the overall transmission time. To achieve this, we allow the preemption of communication tasks. However, in practice preemption comes with a cost, d, and this renders the problem NP-hard (Gopal et al., An optimal switching algorithm for multibeam satellite systems with variable bandwidth beams, IEEE Trans. Commun., vol.30, pp. 2475–2481, 1982). In this paper, we present a approximation algorithm, which is the first one for the PBS problem with approximation ratio strictly less than two. Furthermore, we propose a simple optimal polynomial time algorithm for a subclass of instances of the PBS problem.This work has been partially supported by the the European Union (FET-Working Group APPOL II), and the Greek General Secretariat of Research and Technology.     

Packing trees in communication networks

Given an undirected edge-capacitated graph and given (possibly) different subsets of vertices, we consider the problem of selecting a maximum (weighted) set of Steiner trees, each tree spanning a subset of vertices, without violating the capacity constraints. This problem is motivated by applications in multicast communication networks. We give an integer linear programming (ILP) formulation for the problem, and observe that its linear programming (LP) relaxation is a fractional packing problem with exponentially many variables and a block (sub-)problem that cannot be solved in polynomial time. To this end, we take an r-approximate block solver (a weak block solver) to develop a (1−ε)/r-approximation algorithm for the LP relaxation. The algorithm has a polynomial coordination complexity for any ε∈(0,1). To the best of our knowledge, this is the first approximation result for fractional packing problems with only weak block solvers (with arbitrarily large approximation ratio) and a coordination complexity that is polynomial in the input size. This leads also to an approximation algorithm for the underlying tree packing problem. Finally, we extend our results to an important multicast routing and wavelength assignment problem in optical networks, where each Steiner tree is to be assigned one of a limited set of given wavelengths, so that trees crossing the same fiber are assigned different wavelengths. A preliminary version of this paper appeared in the Proceedings of the 1st Workshop on Internet and Network Economics (WINE 2005), LNCS, vol. 3828, pp. 688–697. Research supported by a MITACS grant for all the authors, an NSERC post doctoral fellowship for the first author, the NSERC Discovery Grant #5-48923 for the second and fourth author, NSERC Discovery Grant #15296 for the third author, the Canada Research Chair Program for the second author, and an NSERC industrial and development fellowship for the fourth author.     

An Improved Randomized Approximation Algorithm for Max TSP

We present an O(n³)-time randomized approximation algorithm for the maximum traveling salesman problem whose expected approximation ratio is asymptotically , where n is the number of vertices in the input (undirected) graph. This improves the previous best.Part of work done while visiting City University of Hong Kong.     

Approximation Algorithms for the Multiple Knapsack Problem with Assignment Restrictions

Motivated by a real world application, we study the multiple knapsack problem with assignment restrictions (MKAR). We are given a set of items, each with a positive real weight, and a set of knapsacks, each with a positive real capacity. In addition, for each item a set of knapsacks that can hold that item is specified. In a feasible assignment of items to knapsacks, each item is assigned to at most one knapsack, assignment restrictions are satisfied, and knapsack capacities are not exceeded. We consider the objectives of maximizing assigned weight and minimizing utilized capacity.We focus on obtaining approximate solutions in polynomial computational time. We show that simple greedy approaches yield 1/3-approximation algorithms for the objective of maximizing assigned weight. We give two different 1/2-approximation algorithms: the first one solves single knapsack problems successively and the second one is based on rounding the LP relaxation solution. For the bicriteria problem of minimizing utilized capacity subject to a minimum requirement on assigned weight, we give an (1/3,2)-approximation algorithm.     

Facility Location with Dynamic Distance Functions

Facility location problems have always been studied with theassumption that the edge lengths in the network are static anddo not change over time. The underlying network could be used to model a city street networkfor emergency facility location/hospitals, or an electronic network for locating information centers. In any case, it is clear that due to trafficcongestion the traversal time on links changes with time. Very often, we have estimates as to how the edge lengths change over time, and our objective is to choose a set of locations (vertices) ascenters, such that at every time instant each vertex has a center close to it (clearly, the center close to a vertex may change over time). We also provide approximation algorithms as well as hardness results forthe K-center problem under this model. This is the first comprehensive study regarding approximation algorithmsfor facility location for good time-invariant solutions.     

The Greedier the Better: An Efficient Algorithm for Approximating Maximum Independent Set

The classical greedy heuristic for approximating maximum independent set is simple and efficient. It achieves a performance ratio of ( + 2)/3, where is the maximum node degree of the input graph. All known algorithms for the problem with better performance ratios are much more complicated and inefficient. In this paper, we propose a natural extension of the greedy heuristic. It is as simple and as efficient as the classical greedy heuristic. By a careful analysis on the structure of the intermediate graphs manipulated by our heuristic, we prove that the performance ratio is improved to ( + 3)/3.25.     

New Approximation Algorithms for Map Labeling with Sliding Labels

In this paper we present approximation algorithm for the following NP-hard map labeling problem: Given a set S of n distinct sites in the plane, one needs to place at each site a uniform square of maximum possible size such that all the squares are along the same direction. This generalizes the classical problem of labeling points with axis-parallel squares and restricts the most general version where the squares can have different orientations. We obtain factor-4 and factor- approximation algorithms for this problem. These algorithms also work for two generalized versions of the problem. We also revisit the problem of labeling each point with maximum uniform axis-parallel square pairs and improve the previous approximation factor of 4 to 3.     

Approximation algorithms for connected facility location problems

We study Connected Facility Location problems. We are given a connected graph G=(V,E) with nonnegative edge cost c _e for each edge e∈E, a set of clients D⊆V such that each client j∈D has positive demand d _j and a set of facilities F⊆V each has nonnegative opening cost f _i and capacity to serve all client demands. The objective is to open a subset of facilities, say , to assign each client j∈D to exactly one open facility i(j) and to connect all open facilities by a Steiner tree T such that the cost is minimized for a given input parameter M≥1. We propose a LP-rounding based 8.29 approximation algorithm which improves the previous bound 8.55 (Swamy and Kumar in Algorithmica, 40:245–269, 2004). We also consider the problem when opening cost of all facilities are equal. In this case we give a 7.0 approximation algorithm.