Approximation Algorithms via Structural Results for Apex-Minor-Free Graphs

  • Authors:
  • Erik D. Demaine;Mohammadtaghi Hajiaghayi;Ken-Ichi Kawarabayashi

  • Affiliations:
  • MIT Computer Science and Artificial Intelligence Laboratory, Cambridge, USA 02139;AT&T Labs -- Research, Florham Park, USA 07932;National Institute for Informatics, Tokyo, Japan 101-8430

  • Venue:
  • ICALP '09 Proceedings of the 36th International Colloquium on Automata, Languages and Programming: Part I
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We develop new structural results for apex-minor-free graphs and show their power by developing two new approximation algorithms. The first is an additive approximation for coloring within 2 of the optimal chromatic number, which is essentially best possible, and generalizes the seminal result by Thomassen [32] for bounded-genus graphs. This result also improves our understanding from an algorithmic point of view of the venerable Hadwiger conjecture about coloring H -minor-free graphs. The second approximation result is a PTAS for unweighted TSP in apex-minor-free graphs, which generalizes PTASs for TSP in planar graphs and bounded-genus graphs [20,2,24,15]. We strengthen the structural results from the seminal Graph Minor Theory of Robertson and Seymour in the case of apex-minor-free graphs, showing that apices can be made adjacent only to vortices if we generalize the notion of vortices to "quasivortices" of bounded treewidth, proving a conjecture from [10]. We show that this structure theorem is a powerful tool for developing algorithms on apex-minor-free graphs, including for the classic problems of coloring and TSP. In particular, we use this theorem to partition the edges of a graph into k pieces, for any k , such that contracting any piece results in a bounded-treewidth graph, generalizing previous similar results for planar graphs [24] and bounded-genus graphs [15]. We also highlight the difficulties in extending our results to general H -minor-free graphs.