Planarization and Acyclic Colorings of Subcubic Claw-Free Graphs

Planarization and Acyclic Colorings of Subcubic Claw-Free Graphs

Planarization and acyclic colorings of subcubic claw-free graphs Christine Cheng!, Eric McDermid!!, and Ichiro Suzuki!!! Department of Computer Science, University of Wisconsin–Milwaukee, Milwaukee, WI 53211, USA ccheng, mcdermid, suzuki @uwm.edu { } Abstract. We study methods of planarizing and acyclically coloring claw-free subcubic graphs. We give a polynomial-time algorithm that, given such a graph G, produces an independent set Q of at most n/6 vertices whose removal from G leaves an induced planar subgraph P (in fact, P has treewidth at most four). We further show the stronger result that in polynomial-time a set of at most n/6 edges can be identified whose removal leaves a planar subgraph (of treewidth at most four). From an approximability point of view, we show that our results imply 6/5- and 9/8-approximation algorithms, respectively, for the (NP-hard) problems of finding a maximum induced planar subgraph and a maximum planar subgraph of a subcubic claw-free graph, respectively. Regarding acyclic colorings, we give a polynomial-time algorithm that finds an optimal acyclic vertex coloring of a subcubic claw-free graph. To our knowledge, this represents the largest known subclass of subcu- bic graphs such that an optimal acyclic vertex coloring can be found in polynomial-time. We show that this bound is tight by proving that the problem is NP-hard for cubic line graphs (and therefore, claw-free graphs) of maximum degree d 4. An interesting corollary to the algo- ≥ rithm that we present is that there are exactly three subcubic claw-free graphs that require four colors to be acyclically colored. For all other such graphs, three colors suffice. 1 Introduction A simple, finite graph G is said to be claw-free if no vertex of G has three pairwise nonadjacent neighbors. It is subcubic if every vertex of G has degree at most three. Claw-free graphs are a well-studied and interesting class of graphs that generalize line graphs. Additionally, claw-free graphs are very well-understood, thanks to a complete structure theorem found by Chudnovsky and Seymour [6]. In this paper we explore methods of planarization and acyclic colorings of claw-free subcubic graphs. ! Supported by NSF award CCF-0830678. !! Supported by NSF award CCF-0830678 and UWM Research Growth Initiative. !!! Supported by UWM Research Growth Initiative. 2 Planarization. Planarization is a broad term used to refer to the process of modifying a graph in order to make it planar (see [14] for a survey). Interest in such techniques arises both for combinatorial and theoretical uses, but also for in- dustrial applications (for example, VLSI and layout problems). In the minimum nonplanar vertex deletion problem (MINVD) we wish to compute a smallest set of vertices Q such that G Q is an induced planar subgraph. The complement problem of MINVD is to find− a maximum induced planar subgraph (MAXIPS), where the size of the solution is the number of vertices of G Q. Analogously, the minimum nonplanar edge deletion problem MINED asks− for a smallest set of edges E! such that G E! is a planar subgraph. The complement of MINED is to find a maximum planar− subgraph (MAXPS), where the size of the solution is the number of edges of G E!. − From a complexity perspective, all of these optimization problems are hard, but some positive results are also known. On the negative side, Yannakakis [18, 19] first showed that MINVD and MINED are NP-hard. Independently, Liu and Geldmacher [15] also proved that MINED is NP-hard. Cˇalinescu et al. [4] improved upon this by showing that MINED and MAXPS are Max SNP-hard. Subsequently, Faria et al. [9–11] (note the two different sets of authors) showed that MINVD and MINED are both Max SNP-hard in cubic graphs, and that MAXIPS and MAXPS are NP-hard in cubic graphs. They also showed that MAXIPS in general graphs admits no polynomial-time approximation algorithm with a fixed ratio unless P = NP. We remark that, although Faria et al. [11] do not explicitly state it, the derived instance in their hardness proof for MAXIPS in cubic graphs has the property that every vertex is in a triangle (a necessary and sufficient condition for claw-freeness in a cubic graph). Hence MAXIPS is NP-hard in claw-free cubic graphs. It is also easy to see that their hardness result for MAXPS in cubic graphs implies that MAXPS is also NP-hard in claw-free cubic graphs: a cubic graph G contains a planar subgraph P of size n k (i.e., − we remove k edges to obtain P ) if and only if the claw-free graph G! obtained by replacing each vertex of G with a triangle has a planar subgraph of size 3n k. − On the positive side, Faria et al. [11] gave a 4/3-approximation algorithm for MAXIPS in subcubic graphs. Edwards and Farr [7] presented a polynomial- time algorithm that, given a graph of maximum degree d, finds an induced planar subgraph of size at least 3n/(d + 1). They later generalized this result to graphs of average degree d (in fact, the induced subgraph found is series- parallel). This can be interpreted as a (d + 1)/3-approximation algorithm, which matches the bound given by Faria et al. when the maximum degree is three. Finally, Cˇalinescu et al. [4] gave a 9/4-approximation for MAXPS in general graphs. We remark that there is much other work on finding subgraphs (induced or otherwise) having a particular property – for example, Cˇalinescu et al. [5] have recently given an approximation for maximum series-parallel subgraphs. Also, different algorithmic approaches have been taken, such as the FPT algorithm of Marx and Schlotter [16] for MINVD. Acyclic colorings. An acyclic coloring of a graph G is a proper vertex coloring of G with the property that no cycle is bicolored. The acyclic coloring problem 3 is to compute an acyclic coloring for G using the fewest number of colors. Aside from its intrinsic interest, researchers are interested in this problem because it has applications in computing the Jacobian and Hessian of sparse matrices [12]. Using probabilistic methods, Alon et al. [1] proved that it is always possible to acyclically color the vertices of a graph of maximum degree d using O(d4/3) colors, and showed that there exists graphs requiring Ω(d4/3/(log d)1/3) colors. For fixed values of d, they showed that a straightforward greedy algorithm uses at most d2 + 1 colors. Several authors have presented polynomial-time algorithms [3, 13, 17] showing that graphs of maximum degree 3, 4, and 5, respectively, admit an acyclic 4-, 5-, and 7-coloring. Note that all of these results are approximation algorithms – they will not, in general, return an acyclic coloring with the fewest number of colors possible. Regarding special cases of graphs, Zhang and Bilka [20] (implicitly) give a polynomial-time algorithm showing that every cubic line graph admits an acyclic coloring using at most three colors – a result that will be useful to us later on. We remark that much attention has also been devoted to the acyclic edge coloring of a graph G (defined analogously to acyclic vertex colorings). For our purposes, we mention only that Alon and Zaks [2] have shown that determining if a cubic graph G admits an acyclic coloring with at most three colors is NP- complete. Our Contribution. On the surface, planarization and acyclic colorings seem to be quite different problems. A unifying approach to our results is our sim- ple reduction technique, that, in linear time, yields a reduced graph with very desirable properties. In particular, we show that if the reduced graph admits a (induced) planar subgraph of a certain size, or is acyclically colorable, then the original graph also has this property. We believe that this reduction technique could be applicable to other problems restricted to subcubic claw-free graphs. Regarding planarization, we give a polynomial-time algorithm that, given a claw-free subcubic graph, produces an independent set Q of at most n/6 ver- tices whose removal from G leaves an induced planar subgraph P (in fact, P has treewidth at most four). Hence, G always has an induced planar subgraph (of treewidth at most four) of size at least 5n/6. Given this result, we further show the stronger result that in polynomial-time a set of at most n/6 edges can be identified whose removal leaves a planar subgraph (of treewidth at most four). From the perspective of the combinatorial bounds given by Edwards and Farr [8], our results show the existence of significantly larger induced planar subgraphs when restricted to subcubic claw-free graphs, rather than general subcubic graphs. From an approximability point of view, we show that our first two bounds give a 6/5- and a 9/8-approximation ratio, respectively, for MAX- IPS and MAXPS. This gives a better performance guarantee for the special case of subcubic claw-free graphs than that of the 4/3-approximation algorithm for MAXIPS in general subcubic graphs, and an improvement for this special case over the general 9/4-approximation algorithm for MAXPS. Next, we give a polynomial-time algorithm that finds an optimal acyclic vertex coloring of a subcubic claw-free graph. To our knowledge, this represents 4 Fig. 1. A claw with base u, a diamond and double diamond with endpoints u and v, and an envelope with an acyclic coloring. the largest subclass of subcubic graphs (or any non-trivially degree bounded class of graphs) such that an optimal acyclic vertex coloring can be found in polynomial-time.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    14 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us