Skip to main content

On b-Matchings and b-Edge Dominating Sets: A 2-Approximation Algorithm for the 4-Edge Dominating Set Problem

  • Conference paper
  • First Online:
Approximation and Online Algorithms (WAOA 2021)

Part of the book series: Lecture Notes in Computer Science ((LNTCS,volume 12982))

Included in the following conference series:

  • 355 Accesses

Abstract

We consider a multiple domination version of the edge dominating set problem, called b-EDS. Here, an edge set \(D\subseteq E\) of minimum cardinality is sought in a given graph \(G=(V,E)\) with a demand vector \(b\in \mathbb {Z}^E\) such that each edge \(e\in E\) is dominated by \(b_e\) edges of D. When a solution D is not allowed to be a multi-set, it is called simple b-EDS. We present a 2-approximation algorithm for simple b-EDS for the case of \(\max _{e\in E}b_e\le 4\). The 2-approximation on general graphs has been known when \(\max _{e\in E}b_e\le 3\), and in the most general cases 8/3 is the best approximation ratio known attainable in polynomial time [2]. Our algorithms are designed based on the most natural LP relaxation of b-EDS and maximal b-matchings (or its generalization).

This work is supported in part by JSPS KAKENHI under Grant Numbers 20K11676 and 17K00013.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 49.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 64.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

References

  1. Baker, B.S.: Approximation algorithms for NP-complete problems on planar graphs. J. Assoc. Comput. Mach. 41(1), 153–180 (1994)

    Article  MathSciNet  Google Scholar 

  2. Berger, A., Fukunaga, T., Nagamochi, H., Parekh, O.: Approximability of the capacitated \(b\)-edge dominating set problem. Theoret. Comput. Sci. 385(1–3), 202–213 (2007)

    Article  MathSciNet  Google Scholar 

  3. Berger, A., Parekh, O.: Linear time algorithms for generalized edge dominating set problems. Algorithmica 50(2), 244–254 (2008)

    Article  MathSciNet  Google Scholar 

  4. Berger, A., Parekh, O.: Erratum to: linear time algorithms for generalized edge dominating set problems. Algorithmica 62(1–2), 633–634 (2012)

    Article  MathSciNet  Google Scholar 

  5. Carr, R., Fujito, T., Konjevod, G., Parekh, O.: A \(2\frac{1}{10}\)-approximation algorithm for a generalization of the weighted edge-dominating set problem. J. Comb. Optim. 5(3), 317–326 (2001)

    Article  MathSciNet  Google Scholar 

  6. Chlebík, M., Chlebíková, J.: Approximation hardness of edge dominating set problems. J. Comb. Optim. 11(3), 279–290 (2006)

    Article  MathSciNet  Google Scholar 

  7. Fujito, T.: On Matchings and b-edge dominating sets: a 2-approximation algorithm for the 3-edge dominating set problem. In: Ravi, R., Gørtz, I.L. (eds.) SWAT 2014. LNCS, vol. 8503, pp. 206–216. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-08404-6_18

    Chapter  MATH  Google Scholar 

  8. Fujito, T., Nagamochi, H.: A 2-approximation algorithm for the minimum weight edge dominating set problem. Disc. Appl. Math. 118(3), 199–207 (2002)

    Article  MathSciNet  Google Scholar 

  9. Horton, J.D., Kilakos, K.: Minimum edge dominating sets. SIAM J. Disc. Math. 6(3), 375–387 (1993)

    Article  MathSciNet  Google Scholar 

  10. Hunt, H.B., Ravi, S.S., Marathe, M.V., Rosenkrantz, D.J., Radhakrishnan, V., Stearns, R.E.: A unified approach to approximation schemes for NP- and PSPACE-hard problems for geometric graphs. In: van Leeuwen, J. (ed.) ESA 1994. LNCS, vol. 855, pp. 424–435. Springer, Heidelberg (1994). https://doi.org/10.1007/BFb0049428

    Chapter  Google Scholar 

  11. Ito, T., Kakimura, N., Kamiyama, N., Kobayashi, Y., Okamoto, Y.: Minimum-cost \(b\)-edge dominating sets on trees. Algorithmica 81(1), 343–366 (2019)

    Article  MathSciNet  Google Scholar 

  12. Mitchell, S., Hedetniemi, S.: Edge domination in trees. In: Proceedings of the Eighth Southeastern Conference on Combinatorics, Graph Theory and Computing, Louisiana State University, Baton Rouge, LA, 1977, pp. 489–509. Congressus Numerantium, No. XIX (1977)

    Google Scholar 

  13. Parekh, O.: Edge dominating and hypomatchable sets. In: Proceedings of 13th SODA, pp. 287–291 (2002)

    Google Scholar 

  14. Schmied, R., Viehmann, C.: Approximating edge dominating set in dense graphs. Theoret. Comput. Sci. 414, 92–99 (2012)

    Article  MathSciNet  Google Scholar 

  15. Srinivasan, A., Madhukar, K., Nagavamsi, P., Rangan, C.P., Chang, M.-S.: Edge domination on bipartite permutation graphs and cotriangulated graphs. Inf. Process. Lett. 56(3), 165–171 (1995)

    Article  MathSciNet  Google Scholar 

  16. Yannakakis, M., Gavril, F.: Edge dominating sets in graphs. SIAM J. Appl. Math. 38(3), 364–372 (1980)

    Article  MathSciNet  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Toshihiro Fujito .

Editor information

Editors and Affiliations

Appendices

A Proof of Lemma 1

Proof

We consider only \(y(\delta (v))\) for \(e=\{u,v\}\not \in M\) (as the same arguments apply to \(y(\delta (u))\)). Moreover, if no edge in \(\delta _M(v)\) is upgraded, \(y(\delta (v))\le 1/2\) (recall Observation 2), and we thus consider the case when at least one in \(\delta _M(v)\) is upgraded.

Let \(\delta _M(v)=\{e_1,e_2,\dots ,e_s\}\) for \(s=d_M(v)\) in what follows, sorted in the ascending order of \(b_e\) (i.e., \(i< j\) implies that \(b_{e_i}\le b_{e_j}\)).

  • Case \(d_M(v)=0\). Clearly \(y(\delta (v))=0\) since \(y_e=0, \forall e\in \delta (v)\).

  • Case \(d_M(v)=1\). Since \(y_{e}\le 1/2\) for any \(e\in M\), \(y(\delta (v))\le 1/2\).

  • Case \(d_M(v)=4\). For each \(e\in \delta _M(v)\), \(b_e=4\), and none in \(\delta _M(v)\) can be upgraded as it is fully dominated by M.

  • Case \(d_M(v)=3\). For each \(e\in \delta _M(v)\), \(b_e\in \{3,4\}\).

    • Subcase \(|\delta _M(v)\cap E_3|=3\). Then, none in \(\delta _M(v)\) can be upgraded as it is fully dominated by M.

    • Subcase \(|\delta _M(v)\cap E_3|=2\). The only edge that can be possibly upgraded is \(e_3\) with \(b_{e_3}=4\). Such an upgrade on \(e_3\) results in \(y_{e_3}=1/2\) while the initial values assigned on \(e_1, e_2\) being nullified at the same time, and hence, \(y(\delta (v))=1/2\).

    • Subcase \(|\delta _M(v)\cap E_3|=1\). The edges in \(\delta _M(v)\) that can be possibly upgraded are only \(e_2 \text { and }e_3\) since \(e_1\) with \(b_{e_1}=3\) is fully dominated by M. Suppose \(e_2\) is upgraded before \(e_3\) is. This means that \(|\delta _M(e_2)|=3\), and when \(e_2\) is upgraded, an edge to be added to D must be chosen form \(\delta (v)\setminus F\) if \(e_3\) is not yet fully dominated, by addition of such an edge, \(e_3\) becomes fully dominated. Therefore, \(e_3\) already is or becomes fully dominated when \(e_2\) is upgraded, and hence, it never happens to upgrade \(e_3\) itself. The upgrading stage also shifts the values of y such that, while resetting \(y_{e_1}\) to 0, \(y_{e_2}\) is increased from initial 1/8 by 1/8 for \(e_1\) and another 1/8 for an edge newly added to D, thus \(y_{e_2}\) totaling to 3/8. Meanwhile, \(y_{e_3}\) remains at the initial value of 1/8, and hence, \(y(\delta (v))=\sum _{i=1}^3y_{e_i}=0+3/8+1/8=1/2\).

    • Subcase \(|\delta _M(v)\cap E_3|=0\). So, \(\delta _M(v)\subseteq E_4\), and if \(e_i\in \delta _M(v)\) is to be upgraded, it must be the case that \(|\delta _M(e_i)|=3\), which means that \(\delta _M(e_i)=\delta _M(v)\). If only one of \(e_1,e_2,e_3\) is to be upgraded, then \(y(\delta (v))=\sum _{i=1}^3y_{e_i}=2/8+1/8+1/8=1/2\) after the upgrade. If two of them are possibly upgradable, then such an upgrade would hinder any subsequent upgrade for the following reason; the first upgrade on e would cause an addition of one edge to D, and that edge must come from \(\delta (v)\setminus F\) because \(|\delta _M(e)|=3\) and \(\delta (v)\) contains other edges not yet fully dominated. So the first upgrade makes all edges in \(\delta _M(v)\) fully dominated. Therefore, at most one out of three is possibly upgraded again in this case, and we have \(y(\delta (v))=\sum _{i=1}^3y_{e_i}=2/8+1/8+1/8=1/2\).

  • Case \(d_M(v)=2\). Let \(\delta _M(v)=\{e_1,e_2\}\) and we know that \(2\le b_{e_1}\le b_{e_2}\le 4\).

    • Subcase \(b_{e_1}=2\): So, \(e_1\) is fully dominated by M (cannot be upgraded), and \(e_2\) must be upgraded, implying that \(b_{e_2}\ge 3\). Such an upgrade would reset \(y_{e_1}\) to 0 while \(y_{e_2}\le 1/2\), and hence, \(y(\delta (v))\le 1/2\).

    • Subcase \(b_{e_1}=3\): We first consider the case when \(e_1\) is upgraded, which implies that \(\delta _M(e_1)=\delta _M(v)\). So, this upgrade chooses an edge from \(\delta (v)\setminus F\) and adds it to D. If \(e_2\in E_3\), \(e_2\) becomes fully dominated as a result of upgrading \(e_1\), and hence, \(y_{e_1}=2/6\text { and }y_{e_2}=1/6\). Suppose now that \(e_2\in E_4\), and after the upgrade on \(e_1\), either \(e_2\) can be fully dominated or not. In the former case, \(y(\delta (v))\) ends up with \(y_{e_1}=2/6\text { and }y_{e_2}=1/8\). In the latter case, \(e_2\) is also upgraded and it causes resetting 2/6 of \(y_{e_1}\) to 0 while increasing \(y_{e_2}\) to no larger than 1/2, and hence, \(y(\delta (v))\) becomes \(\le 1/2\).

      Let us consider next the case when \(e_1\) is not upgraded, implying that upgraded is \(e_2\). In this case again, \(y_{e_1}=1/6\text { and }y_{e_2}=2/6\) if \(e_2\in E_3\), or \(y_{e_1}=0\text { and }y_{e_2}\le 1/2\) if \(e_2\in E_4\), and hence, \(y(\delta (v))\le 1/2\) in either case.

    • Subcase \(b_{e_1}=4\): In this case \(\delta _M(v)\subseteq E_4\). If only one of \(e_1,e_2\) is upgraded, say \(e_2\), then \(y_{e_1}=1/8\) and \(y_{e_2}\) becomes no larger than 3/8, and hence, \(y(\delta (v))\le 1/2\). It remains to consider the case both \(e_1,e_2\) are upgraded, and assume w.l.o.g. that \(|\delta _M(e_1)|\le |\delta _M(e_2)|\) and that \(e_1\) is upgraded before \(e_2\) is.

      • Subsubcase \(|\delta _M(e_1)|=3\): So, \(|\delta _M(e_2)|=3\), and \(y_{e_1}\) becomes 2/8 while \(y_{e_2}\) does at most 2/8, and hence, \(y(\delta (v))\le 1/2\).

      • Subsubcase \(|\delta _M(e_1)|=2\) (or \(\delta _M(e_1)=\delta _M(v)\)): When \(e_1\) is upgraded, two edges are added to D. The algorithm would choose the 1st of those two edges from \(\delta (v)\setminus D\) since \(e_2\), not yet fully dominated edge, is incident with v. Since \(e_2\) is supposed to be still subsequently upgraded, it must be the case that \(\delta _M(e_2)=\delta _M(v)\), and the algorithm would try to choose the 2nd of those two again from \(\delta (v)\setminus D\), if it exists, and if it does, \(e_2\) becomes fully dominated, unnecessitating an upgrade on \(e_2\). Therefore, it is possible for both \(e_1\text { and }e_2\) to be upgraded only when \(\delta _M(e_1)=\delta _M(e_2)=\delta _M(v)\) and \(d(v)=3=d_M(v)+1\), but this is exactly the case when e is an edge of Type 1.

   \(\square \)

B Proof of Lemma 2

Proof

First of all if no edge in \(\delta (e)\) is upgraded, \(y(\delta (e))=y(\delta (u))+y(\delta (v))-y_e\le 1\) because of Observation 2.

Because of Observation 3.2, clearly \(y(\delta (e))\le 1\) if \(|\delta _M(e)|\le 2\). So we consider only cases when \(|\delta _M(e)|\ge 3\).

  • Case \(|\delta _M(e)|=3\). If none in \(\delta _M(e)\) is upgraded, it can be verified that

    $$ y(\delta (e))\le 2\cdot \frac{1}{4}+\frac{1}{6} < 1 . $$

    So, we may assume that at least one in \(\delta _M(e)\) is upgraded.

    • Subcase \(d_M(u)=3, d_M(v)=1\) (or vice versa). For any \(e'\in \delta _M(e)\) we have \(b_{e'}\ge 3\). This means that \(y_{e'}\le 2/8\) even after those edges are upgraded, and hence, \(y(\delta (e))\le 6/8\).

    • Subcase \(d_M(u)=d_M(v)=2\). Let \(\delta _M(u)=\{e,e_1\}\) and \(\delta _M(v)=\{e,e_2\}\). Then, \(b_{e_1},b_{e_2}\ge 2\) and \(b_{e}\ge 3\). This means that \(y_{e_1}, y_{e_2}\le 3/8\) and \(y_e\le 2/8\) even after those edges are upgraded, and hence, \(y(\delta (e))\le 1\).

  • Case \(|\delta _M(e)|=4\). If none in \(\delta _M(e)\) is upgraded, it can be verified that

    $$ y(\delta (e))\le \frac{1}{4}+ \frac{1}{8}+2\cdot \frac{1}{6} < 1 . $$

    So, we may assume that at least one in \(\delta _M(e)\) is upgraded.

    • Subcase \(d_M(u)=4, d_M(v)=1\) (or vice versa). In this case, \(\delta _M(e)\subseteq E_4\), and none can be upgraded.

    • Subcase \(d_M(u)=3, d_M(v)=2\) (or vice versa). Let \(\delta _M(u)=\{e,e_1,e_2\}\) and \(\delta _M(v)=\{e,e_3\}\). If \(e\in E_4\), then it cannot be upgraded while \(y_{e_1}, y_{e_2}\le 2/8\) and \(y_{e_3}\le 3/8\). Hence,

      $$ y(\delta (e))\le \frac{1+2+2+3}{8}=1 $$

      if \(e\in E_4\).

      Consider next the case when \(e\in E_3\), which cannot be upgraded. If \(e_3\in E_3\cup E_2\), then \(y_{e_3}\le 2/6\), and by the same reasoning as above,

      $$ y(\delta (e))\le \frac{2+2}{8}+\frac{1+2}{6}=1 . $$

      So, let us assume that \(e_3\in E_4\). If \(e_3\) is not upgraded, easily \(y(\delta (e))\le 1\), and so suppose it is upgraded. But then, such an operation would absorb 1/6 of \(y_e\) while \(y_{e_3}\le 4/8\). Thus again,

      $$ y(\delta (e))\le \frac{2+2}{8}+0+\frac{4}{8}=1 . $$
  • Case \(|\delta _M(e)|=6\). So, \(d_M(u)=4\text { and }d_M(v)=3\) (or vice versa). Then, \(\delta _M(u)\subseteq E_4\), no edge in it can be upgraded, and \(y(\delta _M(u))=4/8\). Meanwhile, \(y_{e'}\) cannot become larger than 2/8 for each edge \(e'\in \delta _M(v)\setminus \{e\}\). Therefore, \(y(\delta (e))\le (4+2+2)/8=1\).

  • Case \(|\delta _M(e)|=7\). So, \(d_M(u)=d_M(v)=4\), \(\delta _M(e)\subseteq E_4\), and none in \(\delta _M(e)\) can be upgraded.

  • Case \(|\delta _M(e)|=5\). If \(d_M(u)=4\text { and }d_M(v)=2\) (or conversely, \(d_M(u)=2\text { and }d_M(v)=5\)), since every edge in \(\delta _M(u)\) belongs to \(E_4\) and is not upgradable,

    $$ y(\delta (e))\le \frac{4}{8}+\frac{3}{8}=\frac{7}{8} . $$

    It remains to consider the case when \(d_M(u)=d_M(v)=3\). Observe first that \(\delta _M(e)\subseteq E_3\cup E_4\), and e as well as any edge in \(E_3\) is not upgradable. So, only such an edge \(e'\in \delta _M(e)\cap E_4\) with \(|\delta _M(e')|=3\) is upgradable. Let \(\delta _M(u)=\{e,e_1,e_2\}\) and suppose \(e_1\) is such an edge and the first edge upgraded within \(\delta _M(e)\). Observe here that, if \(e_2\in E_3\) or if \(d(u)>d_M(u)\), \(e_2\) cannot be upgraded after \(e_1\) is, and when \(e_1\) is upgraded, \(y(\delta (u))\) becomes 4/8 whether \(e, e_2\) are in \(E_3\) or in \(E_4\). So, in this case, \(y(\delta (e))\le y(\delta (u))+2/8+2/8= 1\).

    Similarly, if any in \(\delta _M(v)\setminus \{e\}\) is not upgraded, \(y(\delta (e))\le 1\). Thus, the only remaining case is when all the edges in \(\delta _M(e)\) except for e are upgraded, and this is possible if and only if \(d(u)=d(v)=3\), \(e'\in E_4\), and \(|\delta _M(e')|=3\) for all edges \(e'\in \delta _M(e)\setminus \{e\}\), and this is exactly when e is an edge of Type 2.

Thus, we may conclude that \(y(\delta (e))>1\) for \(e\in M\) only if e is an edge of Type 2.

   \(\square \)

Rights and permissions

Reprints and permissions

Copyright information

© 2021 Springer Nature Switzerland AG

About this paper

Check for updates. Verify currency and authenticity via CrossMark

Cite this paper

Fujito, T., Tatematsu, T. (2021). On b-Matchings and b-Edge Dominating Sets: A 2-Approximation Algorithm for the 4-Edge Dominating Set Problem. In: Koenemann, J., Peis, B. (eds) Approximation and Online Algorithms. WAOA 2021. Lecture Notes in Computer Science(), vol 12982. Springer, Cham. https://doi.org/10.1007/978-3-030-92702-8_5

Download citation

  • DOI: https://doi.org/10.1007/978-3-030-92702-8_5

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-030-92701-1

  • Online ISBN: 978-3-030-92702-8

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics