On the hull number of some graph classes

On the hull number of some graph classes Julio Araújo, Victor Campos, Frédéric Giroire, Leonardo Sampaio, Ronan Pardo Soares To cite this version: Julio Araújo, Victor Campos, Frédéric Giroire, Leonardo Sampaio, Ronan Pardo Soares. On the hull number of some graph classes. European Conference on Combinatorics, Graph Theory and Applications (EuroComb’11), Rényi Institute, Aug 2011, Budapest, Hungary. pp.49-55, ฀10.1016/j.endm.2011.09.009฀. ฀inria-00635032฀ HAL Id: inria-00635032 https://hal.inria.fr/inria-00635032 Submitted on 24 Oct 2011 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. On the hull number of some graph classes. 1 J. Araujo a,b V. Campos b F. Giroire a L. Sampaio a,2 R. Soares a,b a {julio.araujo, frederic.giroire, leonardo.sampaio rocha, ronan.pardo soares}@inria.fr - MASCOTTE Project, I3S (CNRS & UNS) and INRIA, INRIA Sophia-Antipolis, 2004 route des Lucioles, BP 93, 06902 Sophia-Antipolis Cedex France. b {[email protected]} ParGO Research Group - Universidade Federal do Ceará UFC - Campus do Pici, Bloco 910. 60455-760 - Fortaleza, CE - Brazil. Abstract Given a graph G = (V, E), the closed interval of a pair of vertices u, v ∈ V , denoted by I[u, v], is the set ofSvertices that belongs to some shortest (u, v)-path. For a given S ⊆ V , let I[S] = u,v∈S I[u, v]. We say that S ⊆ V is a convex set if I[S] = S. The convex hull Ih [S] of a subset S ⊆ V is the smallest convex set that contains S. We say that S is a hull set if Ih [S] = V . The cardinality of a minimum hull set of G is the hull number of G, denoted by hn(G). We show that deciding if hn(G) ≤ k is an NP-complete problem, even if G is bipartite. We also prove that hn(G) can be computed in polynomial time for cactus and P4 -sparse graphs. Keywords: graph convexity, hull number, bipartite graph, cactus graph, P4 -sparse graph. 1 Introduction In this paper, we adopt the graph terminology defined in [2]. All the graphs studied in this work are considered to be simple and undirected. 1 2 Supported by the INRIA associated team EWIN. Partially supported by CAPES/Brazil and ANR Blanc AGAPE. Given a graph G = (V, E), the closed interval of a pair of vertices u, v ∈ V , denoted by I[u, v], is the set S of vertices that belong to some shortest (u, v)path. If S ⊆ V , let I[S] = u,v∈S I[u, v]. We say that S ⊆ V is a convex set if I[S] = S. Given a subset S ⊆ V , the convex hull Ih [S] of S is the smallest convex set that contains S. We say that S is a hull set if Ih [S] = V . The cardinality of a minimum hull set of G is the hull number of G, denoted by hn(G). The Hull Number problem is to determine whether hn(G) ≤ k, for a given graph G and an integer k. This problem is NP-complete [5]. The Hull Number problem was defined in [7]. In [8], results of abstract convexity are shown to be valid for graph convexities. An oriented version of the Hull Number problem is studied in [4]. The hull number of a cartesian product of two connected graphs is characterized in [3]. Dourado et al. [6] show some bounds for the hull number of triangle-free graphs. In [5], it is also proved that the hull number of unit interval graphs, cographs and split graphs can be computed in polynomial time. In Section 3, we answer an open question of Dourado et al. [5] by showing that it is NP-hard to compute the hull number of a given bipartite graph. We then present polynomial time algorithms to compute the hull number of cacti, in Section 4, and of P4 -sparse graphs in Section 5. This last result extends the previous known result for cographs [5]. Finally, we present some concluding remarks. 2 Preliminaries In order to present our main results, it is necessary to first point out some observations on hull sets. Lemma 2.1 ([5]) Every simplicial vertex of a graph G belongs to any hull set of G. Lemma 2.2 ([5]) If a graph G is not complete, then every universal vertex u of G does not belong to any minimum hull set of G. Lemma 2.3 ([5]) Let G be a graph and H be an isometric subgraph of G. Then for every hull set S of H it holds that V (H) ⊆ Ih (S)G . Lemma 2.4 ([5]) Let G be a graph and S a proper and non-empty subset of V (G). If V (G)\S is convex, then every hull set of G contains at least one vertex of S. Fig. 1. Subgraph of the bipartite instance G(F) containing the gadget of a variable xi that appears positively in clauses C1 and C2 , and negatively in C8 . If xi appears positively in Cj , link a6i to cj through yij . If it appears negatively, we use b6i instead of a6i . See the appendix for more details. 3 Bipartite graphs In this section, we study the complexity of the Hull Number problem for bipartite graphs. Theorem 3.1 Given a bipartite graph G and an integer k, deciding if hn(G) ≤ k is an NP-complete problem. Sketch of the proof: This proof is a modification of the one presented for general graphs [5]. We reduce SAT to Hull Number in bipartite graphs by creating an instance as depicted in Figure 1. As it happens in the proof in [5]: a given SAT formula F is satisfied if, and only if, the hull number of the corresponding bipartite instance is at most n + 1, where n is the number of variables of F. ✷ Now, we show some approximability results. Let IG(G) be the incidence graph of G, obtained from G by adding one vertex suv , for each edge uv ∈ E(G), and replacing the edge uv by the edges usuv , suv v. Proposition 3.2 hn(IG(G)) ≤ hn(G) ≤ 2hn(IG(G)). Corollary 3.3 If there exists a k-approximation algorithm B to compute the hull number of bipartite graphs, then B is a 2k-approximation algorithm for any graph. 4 Cacti graphs It is easy to observe that the hull number of a tree is equal to its number of leaves. In this section, we generalize this simple result to the class of cacti. A cactus is a graph in which every pair of cycles have at most one common vertex. In order to compute the hull number of a cactus, we show how to select the vertices of a minimum hull set. Lemma 4.1 Let C be an induced cycle in G. If C has at most one vertex v with degree strictly more than 2, then any minimum hull set of G contains either one or two vertices of V (C). Lemma 4.2 Let C be an induced cycle in G. Let v1 , . . . , vk be the vertices of C having degree greater than two, for k ≥ 2. If there is a set of vertices S of C such that they are not in any shortest (vi , vj )-path, 1 ≤ i, j ≤ k, in C, then any minimum hull set of G contains exactly one vertex of S. Theorem 4.3 Given a cactus G, hn(G) can be found in linear time. Proof. By a simple breadth-first search all the cycles of a cactus can be detected. For each one, it is necessary to detect if it has at most one vertex of degree greater than two. In case it has at least two vertices of degree greater than two, it is necessary to also verify if there is a set S satisfying Lemma 4.2. Finally, every vertex of degree one must be chosen. These tests can be done in linear time and the result follows by Lemmas 2.1, 4.1 and 4.2. ✷ 5 P4 -sparse graphs A graph is said to be a cograph if it is P4 -free. Computing the hull number of cographs can be done in polynomial time [5]. In this section, we extend this result to the class of P4 -sparse graphs which properly contains the cographs. A graph is P4 -sparse if, for every S ⊆ V such that |S| = 5, G[S] contains at most one induced P4 . Clearly, a cograph is a P4 -sparse graph. We call G a strict P4 -sparse graph if G is P4 -sparse and contains at least one P4 as induced subgraph. The algorithm we present to compute the hull number of any P4 -sparse graph G uses its modular decomposition tree T (G). The idea is to perform a post-order traversal in T (G) calculating hn(G). The definition and classical results about the modular decomposition of graphs can be found in [10]. It is known that the neighborhood nodes of the modular decomposition of P4 -sparse graphs are spiders [9]. A spider is a graph whose vertex set can be partitioned into disjoint sets S, K, and R such that: |S| = |K| ≥ 2, S is a stable set, K is a clique; every vertex in R is adjacent to all the vertices in K and to no vertex in S; there exists a bijection f : S −→ K such that, for all vertices s ∈ S, either NG (s) ∩ K = K − {f (s)} (thick spider) or NG (s) ∩ K = {f (s)} (thin spider). Note that R may be empty, it is the only possible non-trivial strong submodule of a spider, and that if |S| = |K| = 2, the spider is, by definition, thin and thick. However, in order to simplify the notation in the proofs, we consider that a spider is called thick if, and only if, |S| = |K| ≥ 3. Moreover, we remark that it is possible to find a modular decomposition tree in such a way that the series and parallel nodes of the tree have only two children each. Finally, we present how to compute the hull number of each inner node of the modular decomposition tree of a P4 -sparse graph. First, we need the following lemma: Lemma 5.1 Let G = (V, E) be an arbitrary graph such that V can be partitioned into three sets: a clique K, a non-empty set R which is not a clique, and S = V \(K ∪ R) with the property that, for every vertex v ∈ R, we have (N (v) ∩ (V \R)) = K. Then any minimum hull set of G contains a minimum hull set of the subgraph H ⊆ G, induced by K ∪ R. Lemma 5.2 ([5]) If G = G1 ∪ G2 , for any two graphs G1 and G2 , then hn(G) = hn(G1 ) + hn(G2 ). Lemma 5.3 Let G1 and G2 be two P4 -sparse graphs and G = G1 ⊕G2 . Then, hn(G) can be computed in polynomial-time. Lemma 5.4 (Let G = (S ∪ K ∪ R, E) be a spider. Then, |S| + hn(G[K] ⊕ G[R]), if R 6= ∅ and R is not a clique; hn(G) = |S| + |R|, otherwise. Theorem 5.5 Given a P4 -sparse graph G, the hull number of G can be found in polynomial time. Proof. The modular decomposition tree of any graph G can be found in linear time [11]. The complexity of the post-order traversal is O(n). For each inner node, the algorithm may spend O(n) operations in case 2 of Lemma 5.3 or when R 6= ∅ in Lemma 5.4. Consequently, the complexity of the algorithm is O(n2 ). ✷ 6 Conclusions In this paper, we simplified the reduction of Dourado et al. [5] to answer two questions they asked about the complexity of computing the hull number of triangle-free graphs and, in particular, of bipartite graphs. However, the complexity of this problem for interval graphs and chordal graphs remains open. Finally, we used the modular decomposition tree of P4 -sparse graphs to compute the hull number of these graphs. The P4 -sparse graphs are contained in the class of (q, q − 4)-graphs for which every set of q vertices induce at most q − 4 disjoint P4 ’s [1]. It would be nice to extend our results to this class of graphs. References [1] Babel, L. and S. Olariu, On the structure of graphs with few p4s, Discrete Appl. Math. 84 (1998), pp. 1–13. [2] Bondy, J. A. and U. S. R. Murty, “Graph Theory,” Graduate Texts in Mathematics, Springer, 2008. [3] Cagaanan, G. B., S. R. Canoy and Jr., On the hull sets and hull number of the cartesian product of graphs, Discrete Mathematics 287 (2004), pp. 141 – 144. [4] Chartrand, G., J. F. Fink and P. Zhang, The hull number of an oriented graph, International Journal of Mathematics and Mathematical Sciences 2003 (2003), pp. 2265–2275. [5] Dourado, M. C., J. G. Gimbel, J. Kratochvı́l, F. Protti and J. L. Szwarcfiter, On the computation of the hull number of a graph, Discrete Mathematics 309 (2009), pp. 5668 – 5674, combinatorics 2006, A Meeting in Celebration of Pavol Hell’s 60th Birthday (May 1-5, 2006). [6] Dourado, M. C., F. Protti, D. Rautenbach and J. L. Szwarcfiter, On the hull number of triangle-free graphs, SIAM J. Discret. Math. 23 (2010), pp. 2163– 2172. [7] Everett, M. G. and S. B. Seidman, The hull number of a graph, Discrete Mathematics 57 (1985), pp. 217 – 223. [8] Farber, M. and R. E. Jamison, Convexity in graphs and hypergraphs, SIAM J. Algebraic Discrete Methods 7 (1986), pp. 433–444. [9] Giakoumakis, V. and J. Vanherpe, On extended P4 -reducible and extended P4 sparse graphs, Theoretical Computer Science 180 (1997), pp. 269–286. [10] McConnell, R. M. and J. P. Spinrad, Modular decomposition and transitive orientation, Discrete Mathematics 201 (1999), pp. 189–241. [11] Tedder, M., D. G. Corneil, M. Habib and C. Paul, Simpler Linear-Time Modular Decomposition Via Recursive Factorizing Permutations, in: Proceedings of the 35th International Colloquium on Automata, Languages and Programming (ICALP), Lecture Notes in Computer Science 5125, 2008, pp. 634–645.