Operations Research Letters 37 (2009) 181–186 Contents lists available at ScienceDirect Operations Research Letters journal homepage: www.elsevier.com/locate/orl An edge-reduction algorithm for the vertex cover problem Qiaoming Han a , Abraham P. Punnen b, , Yinyu Ye c a School of Mathematics and Statistics, Zhejiang University of Finance and Economics, Hangzhou 310018, PR China b Department of Mathematics, Simon Fraser University, 14th Floor Central City Tower, 13450 102nd Ave., Surrey, BC V3T5X3, Canada c Department of Management Science and Engineering, School of Engineering, Stanford University, Stanford, CA 94305, USA article info Article history: Received 19 April 2008 Accepted 20 January 2009 Available online 3 February 2009 Keywords: Vertex cover problem Approximation algorithm LP-relaxation abstract An approximation algorithm for the vertex cover problem is proposed with performance ratio 3 2 on special graphs. On an arbitrary graph, the algorithm guarantees a vertex cover S 1 such that |S 1 |≤ 3 2 |S |+ξ where S is an optimal cover and ξ is an error bound identified. © 2009 Elsevier B.V. All rights reserved. 1. Introduction Let G = (V , E ) be an undirected graph on the vertex set V = {1, 2,..., n}.A vertex cover of G is a subset S of V such that each edge of G has at least one endpoint in S . The vertex cover problem (VCP) is to compute a vertex cover of smallest cardinality in G. VCP is a classical NP-hard problem. It is well known that an optimal vertex cover of a graph can be approximated within a factor of 2 in polynomial time by taking all the vertices of a maximal (not necessarily maximum) matching in the graph or rounding the LP relaxation solution of an integer programming formulation [1,15]. There has been considerable work (see e.g. survey paper [2]) on the problem over the past 30 years on finding a polynomial-time approximation algorithm with an improved performance guarantee. It is known that computing a δ-approximate solution in polynomial time for the VCP is NP-Hard for any δ 10 5 21 1.36 [3]. In fact, no polynomial-time (2 ǫ)-approximation algorithm is known for the VCP for any constant ǫ > 0. Under the assumption of unique game conjecture [4–6] is valid, many researchers believe that a polynomial time 2 ǫ approximation algorithm is not possible for the VCP for any constant ǫ > 0. The current best known bound on the performance ratio of a polynomial time approximation algorithm for VCP is 2 Θ( 1 log n ) [7]. Halperin [8] showed that an approximation ratio of 2 2 log log log can be obtained with the semidefinite programming (SDP) relaxation of VCP where Corresponding author. E-mail addresses: qmhan@nju.edu.cn (Q. Han), apunnen@sfu.ca (A.P. Punnen), yinyu-ye@stanford.edu (Y. Ye). is the maximum degree of G. Other SDP-relaxations of the VCP were studied in [9,10]. On four colorable graphs, a 3 2 -approximate solution can be identified in polynomial time [11]. Recently Asgeirsson and Stein [12,13] reported extensive experimental results using a heuristic algorithm which obtained no worse than 3 2 -approximate solutions for all the test problems they considered. A natural integer programming formulation of VCP can be described as follows: (VC ) min n i=1 x i s.t . x i + x j 1, (i, j) E , x i ∈{0, 1}, i = 1, 2,..., n. (1) Let ¯ x = ( ¯ x 1 , ¯ x 2 ,..., ¯ x n ) be an optimal solution to (1). Then R = {i x i = 1} is an optimal vertex cover of the graph G. The linear programming relaxation of the above integer program, denoted by LP, is given by relaxing the integrality constraints to x i 0, i = 1, 2,..., n. Any vertex cover must contain at least s + 1 vertices of an odd cycle of length 2s + 1. This motivates the following extended linear programming (ELP) relaxation of the VCP: (ELP ) min n i=1 x i s.t . x i + x j 1, (i, j) E , iω k x i s k + 1, ω k , x i 0, i = 1, 2,..., n, (2) where denotes the set of all odd-cycles of G and ω k contains 2s k + 1 vertices for some integer s k . Note that even 0167-6377/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.orl.2009.01.010