# On the Parameterized Complexity of k-Edge Colouring

For every fixed integer k ≥ 1, we prove that k-Edge Colouring is fixed-parameter-tractable when parameterized by the number of vertices of maximum degree.

## Authors

• 14 publications
• 14 publications
• 43 publications
• 15 publications
• ### Parameterized Complexity of Graph Burning

Graph Burning asks, given a graph G = (V,E) and an integer k, whether th...
07/17/2020 ∙ by Yasuaki Kobayashi, et al. ∙ 0

• ### Levelable Sets and the Algebraic Structure of Parameterizations

Asking which sets are fixed-parameter tractable for a given parameteriza...
09/14/2017 ∙ by Jouke Witteveen, et al. ∙ 0

• ### Mixed Integer Programming with Convex/Concave Constraints: Fixed-Parameter Tractability and Applications to Multicovering and Voting

A classic result of Lenstra [Math. Oper. Res. 1983] says that an integer...
09/08/2017 ∙ by Robert Bredereck, et al. ∙ 0

• ### Fixed-parameter tractability of counting small minimum (S,T)-cuts

The parameterized complexity of counting minimum cuts stands as a natura...
07/04/2019 ∙ by Pierre Bergé, et al. ∙ 0

• ### On the parameterized complexity of manipulating Top Trading Cycles

We study the problem of exchange when 1) agents are endowed with heterog...
03/06/2018 ∙ by William Phan, et al. ∙ 0

• ### On the Parameterized Complexity of Maximum Degree Contraction Problem

In the Maximum Degree Contraction problem, input is a graph G on n verti...
09/24/2020 ∙ by Saket Saurabh, et al. ∙ 0

• ### The Longest Run Subsequence Problem: Further Complexity Results

Longest Run Subsequence is a problem introduced recently in the context ...
11/16/2020 ∙ by Riccardo Dondi, et al. ∙ 0

##### This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

## 1 Introduction

For an integer , a -edge colouring of a graph is a mapping such that for any two edges and of that have a common end-vertex. The Edge Colouring problem is to decide if a given graph has a -edge colouring for some given integer . If is fixed, that is, not part of the input, then we denote the problem as:

.99 -Edge Colouring

[2pt]     Instance: a graph . Question: does have a -edge colouring?

The chromatic index of a graph is the smallest integer  such that has a -edge colouring. The degree of a vertex  is the size of its neighbourhood . We let denote the maximum degree of . Vizing showed the following classical result.

###### Theorem 1.1 ([27])

The chromatic index of a graph is either or .

Due to Theorem 1.1, we can make the following observation.

###### Observation 1

Let be a graph. If , then is a yes-instance of -Edge Colouring. If , then is a no-instance of -Edge Colouring.

By Observation 1 we may assume without loss of generality that an input graph of -Edge Colouring has maximum degree . The following well-known hardness result was proven by Holyer for and by Leven and Galil for .

###### Theorem 1.2 ([17, 18])

For every , -Edge Colouring is NP-complete even for graphs in which every vertex has degree .

In this note we consider the -Edge Colouring problem from the viewpoint of Parameterized Complexity, where problem inputs are specified by a main part of size  and a parameter , which is assumed to be small compared to . The main question is to determine if a problem is fixed-parameter tractabe (FPT), that is, if it can be solved in time , where is a computable (and possibly exponential) function of . The choice of parameter depends on the context. By the above discussion, the most natural choice of parameter for -Edge Colouring is the number of vertices of maximum degree. We prove the following result (note that -Edge Colouring is polynomial-time solvable for ).

###### Theorem 1.3

For every , -Edge Colouring can be solved in time on graphs with vertices, of which have maximum degree, and edges. Moreover, it is possible to find a -edge colouring of in time (if it exists).

As we assume that is a fixed constant, Theorem 1.3 implies that for every , the -Edge Colouring problem is fixed-parameter tractable when parameterized by the number of vertices of maximum degree. We prove Theorem 1.3 in Section 2 by using the alternative proof of Theorem 1.1 given by Ehrenfeucht, Faber and Kierstead [13]. We first discuss some related work.

### Related Work

Apart from a number of (classical) complexity results (e.g. [1, 2, 11, 12, 15, 21, 22, 23, 26]), most results on edge colouring are related to Theorem 1.1 and are of a more structural nature. That is, they involve the derivation of sufficient or necessary conditions for a graph to be -edge colourable (see, for example [4, 8, 16, 19]), in which case the graph is said to be Class 1 (and Class 2 otherwise). This is also the focus of papers on edge colouring related to the number of maximum-degree vertices (see, for example, [5, 10, 25, 28]). In particular, Fournier [14] proved that every graph  in which the vertices of degree  induce a forest is Class 1. As a consequence, a graph with at most two vertices of maximum degree is Class 1. In [6] and [7], Chetwynd and Hilton gave necessary and sufficient polynomial-time verifiable conditions for a graph with three, respectively, four vertices of maximum degree to be Class 1. The same authors proved an analogous result for -vertex graphs with vertices of maximum degree in [9], assuming .

For more on edge colouring we refer to the recent survey of Cao, Chen, Jing, Stiebitz and Toft [3].

## 2 The Proof of Theorem 1.3

In this section we prove Theorem 1.3. Let be a graph and let . The graph is the subgraph of induced by . The set is the set of vertices adjacent to at least one vertex in . Let be the set of maximum-degree vertices of . In this context, the graph  is said to be the core of , whereas the graph is said to be the semi-core of .

Machado and de Figueiredo [20] proved the following result.

###### Theorem 2.1 ([20])

Let be a graph and be an integer. Then has a -edge colouring if and only if its semi-core has a -edge colouring.

The proof of Theorem 2.1 is based on an application of the recolouring procedure of Vizing [27] for proving Theorem 1.1 and which was also used by Misra and Gries [24] for giving a constructive proof of Theorem 1.1. As explained by Zatesko [29], the proof of Theorem 2.1 immediately yields a polynomial-time algorithm for finding a -edge colouring of a graph given a -edge colouring of its semi-core (if it exists). Below we give a new, short algorithmic proof of Theorem 2.1, with the same time complexity, via a modification of the alternative proof of Theorem 1.1 by Ehrenfeucht et al. [13], which might be of independent interest. We state this result as a lemma, as we use it to prove our main result.

###### Lemma 1

Let be a graph with a core of size . Given a -edge colouring of its semi-core, it is possible to construct in time a -edge colouring of .

###### Proof

Recall that denotes the set of vertices of maximum degree  of . Let be a -edge colouring of the semi-core of and denote by the size of the semi-core of . Note that . If , then is a -edge colouring of . Assume that .

We write . We let and for . Note that .

We define . We now show how to extend vertex by vertex until we obtain a -edge colouring of the whole graph. That is, for , we show how to construct a -edge colouring  of given a -edge colouring  of

Thus suppose that and suppose that we already have a -edge colouring of (note that at the start of the procedure, when , this is indeed the case, as we have ). We say that a vertex misses colour if none of the edges incident to is coloured by . We denote the set of colours that misses by . Note that . We write . We now iteratively colour the edges incident to in in such a way that at any time, the resulting mapping is a partial -edge colouring of that satisfies the two following properties:

• for each uncoloured edge in , we have ;

• there is at most one uncoloured edge in with .

Observe that at the start of this process, when no edge in has been coloured, satisfies (1) and (2). This can be seen as follows. Vertex is in only adjacent to vertices of (where we define if ). Every that is adjacent to does not belong to and thus has degree at most . Hence, is adjacent to at most vertices in and thus has . As , this means that each edge in has , implying that (1) and (2) are satisfied.

We will now describe how we can maintain properties (1) and (2) while colouring the edges incident to in one by one. Throughout this process we maintain a set , so at the start consists of all neighbours of in . We distinguish two cases.

Case 1. There exists a colour such that appears in at most one set for some with . Then we assign colour to the edge for which is the smallest over all sets that contain . If all these sets have size at least 3, then afterwards (1) and (2) are satisfied. Otherwise, there is a unique smallest set of size at most 2, which also implies that afterwards (1) and (2) are satisfied.

Case 2. For each colour , there exists at least two distinct vertices such that and both and have size at most 2. This means that

 2∣∣ ∣∣⋃v∈WF(uiv)∣∣ ∣∣≤∑v∈W:|F(uiv)|≤2|F(uiv)|≤2|W|.

Hence, we deduce that . As has degree at most  in , it follows that in , at most edges incident to have already been coloured in this stage. It follows that . Hence, there exists a colour . Consequently, each vertex in must have an incident edge coloured . Now choose a vertex for which is minimum and consider a colour . We swap colours and along the path  in that starts in and whose edges are alternatingly coloured and . This yields another partial -edge colouring of . However, now has no incident edge coloured  anymore, which means that we can colour the edge with colour . We observe that for all , the set remains unchanged except the end-vertex of if is adjacent to ; in that case is replaced by . However, by minimality of , we have that and property (2) implies that if , then . Thus, . We conclude that both (1) and (2) are satisfied by the newly obtained (partial) -edge colouring of .

After colouring the last uncoloured edge in incident to  we have obtained our -edge colouring  of . Hence, after doing this for , we have indeed extended to a -edge colouring  of .

Finally note that since , the algorithm has at most extension steps and as each takes time, the total running time is . This completes the proof of Lemma 1. ∎

We are now ready to prove Theorem 1.3, which we restate below.

Theorem 1.3. For every , -Edge Colouring can be solved in time on graphs with vertices, of which have maximum degree, and edges. Moreover, it is possible to find a -edge colouring of in time (if it exists).

###### Proof

Let be an instance of -Edge Colouring. We first compute the maximum degree  of and the corresponding set  of vertices of degree  in time (so . Let denote the semi-core of . By Theorem 2.1, it suffices to check if has a -edge colouring. Note that we may assume by Observation 1, hence . This means that

 2|E(H)|=∑u∈XdH(u)+∑u∈N(X)dH(u)≤∑u∈XdG(u)+∑u∈N(X)dG(u)≤kp+kp(k−1).

Hence, has at most edges. Checking if a mapping is a -edge colouring takes time. The number of such mappings is at most . Hence, brute force checking if has a -edge colouring takes time . Thus the first statement of the theorem follows. We obtain the second statement by applying Lemma 1. ∎

## 3 Conclusions

We proved that -Edge Colouring is fixed-parameter-tractable when parameterized by the number of maximum-degree vertices. We note that our proof does not work to show this for Edge Colouring, as the set may have size . As such, we pose the following open problem: what is the computational complexity of Edge Colouring when parameterized by the number of maximum-degree vertices?

Acknowledgements. We thank Leandro Zatesko for bringing paper [20] to our attention.

## References

• [1] H. L. Bodlaender. Polynomial algorithms for graph isomorphism and chromatic index on partial -trees. Journal of Algorithms, 11(4):631–643, 1990.
• [2] L. Cai and J. A. Ellis. NP-completeness of edge-colouring some restricted graphs. Discrete Applied Mathematics, 30(1):15–27, 1991.
• [3] Y. Cao, G. Chen, G. Jing, M. Stiebitz, and B. Toft. Graph edge coloring: A survey. Graphs and Combinatorics, to appear.
• [4] B.-L. Chen, H.-L. Fu, and M. T. Ko. Total chromatic number and chromatic index of split graphs. Journal of Combinatorial Mathematics and Combinatorial Computing, 17:137–146, 1995.
• [5] A. G. Chetwynd and A. J. W. Hilton. The chromatic index of graphs of even order with many edges. Journal of Graph Theory, 8(4):463–470, 1984.
• [6] A. G. Chetwynd and A. J. W. Hilton. The chromatic index of graphs with at most four vertices of maximum degree. Congressus Numerantium, 43:221–248, 1984.
• [7] A. G. Chetwynd and A. J. W. Hilton. Regular graphs of high degree are -factorizable. Proceedings of the London Mathematical Society, 50:193–206, 1985.
• [8] A. G. Chetwynd and A. J. W. Hilton. A delta-subgraph condition for a graph to be class 1. J. Comb. Theory, Ser. B, 46(1):37–45, 1989.
• [9] A. G. Chetwynd and A. J. W. Hilton. The chromatic index of graphs with large maximum degree, where the number of vertices of maximum degree is relatively small. Journal of Combinatorial Theory, Series B, 48(1):45–66, 1990.
• [10] K. H. Chew. The chromatic index of graphs of high maximum degree. Discrete Mathematics, 162(1-3):77–91, 1996.
• [11] S. M. de Almeida, C. P. de Mello, and A. Morgana. Edge-coloring of split graphs. Ars Combinatoria, 119:363–375, 2015.
• [12] C. de Simone and A. Galluccio. Edge-colouring of regular graphs of large degree. Theor. Comput. Sci., 389(1-2):91–99, 2007.
• [13] A. Ehrenfeucht, V. Faber, and H. Kierstead. A new method of proving theorems on chromatic index. Discrete Mathematics, 52(2):159 – 164, 1984.
• [14] J. Fournier. Méthode et théorème générale de coloration des arêtes. Journal de Mathématiques Pures et Appliquées, 56:437–453, 1977.
• [15] E. Galby, P. T. Lima, D. Paulusma, and B. Ries. Classifying -edge colouring for -free graphs. CoRR, abs/1810.04379, 2018.
• [16] A. J. W. Hilton and C. Zhao. A sufficient condition for a regular graph to be class 1. Journal of Graph Theory, 17(6):701–712, 1993.
• [17] I. Holyer. The NP-completeness of edge-coloring. SIAM Journal on Computing, 10(4):718–720, 1981.
• [18] D. Leven and Z. Galil. NP completeness of finding the chromatic index of regular graphs. Journal of Algorithms, 4(1):35–44, 1983.
• [19] A. R. C. Lima, G. Garcia, L. Zatesko, and S. M. de Almeida. On the chromatic index of cographs and join graphs. Eletronic Notes in Discrete Mathematics, 50:433–438, 2015.
• [20] R. C. S. Machado and C. M. H. de Figueiredo. Decompositions for edge-coloring join graphs and cobipartite graphs. Discrete Applied Mathematics, 158:1336–1342, 2010.
• [21] R. C. S. Machado, C. M. H. de Figueiredo, and N. Trotignon. Edge-colouring and total-colouring chordless graphs. Discrete Mathematics, 313(14):1547–1552, 2013.
• [22] D. S. Malyshev. The complexity of the edge 3-colorability problem for graphs without two induced fragments each on at most six vertices. Sib. elektr. matem. izv., 11:811–822, 2014.
• [23] D. S. Malyshev. Complexity classification of the edge coloring problem for a family of graph classes. Discrete Mathematics and Applications, 27:97–101, 2017.
• [24] J. Misra and D. Gries. A Constructive Proof of Vizing’s Theorem. Information Processing Letters 41(3):131–133, 1992.
• [25] T. Niessen and L. Volkmann. Class 1 conditions depending on the minimum degree and the number of vertices of maximum degree. Journal of Graph Theory, 14(2):225–246, 1990.
• [26] C. Ortiz, N. Maculan, and J. L. Szwarcfiter. Characterizing and edge-colouring split-indifference graphs. Discrete Applied Mathematics, 82(1-3):209–217, 1998.
• [27] V. G. Vizing.

On an estimate of the chromatic class of a

-graph.
Diskret. Analiz., 3:25–30, 1964.
• [28] H. P. Yap. Some Topics in Graph Theory. Cambridge University Press, New York, NY, USA, 1986.
• [29] L.M. Zatesko. Novel Procedures for Graph Edge-Coloring, PhD Thesis, Federal University of Paraná, 2018.