Approximation schemes via Width/Weight trade-offs on minor-free graphs

Fedor V. Fomin, Daniel Lokshtanov, Saket Saurabh, Meirav Zehavi

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Scopus citations


In this paper, we prove a new scaling lemma for vertex weighted minor free graphs that allows for a smooth trade-off between the weight of a vertex set S and the treewidth of G − S. More precisely, we show the following. There exists an algorithm that given an Hminor free graph G, a weight function w : V (G) → Q+ and integers t and s, runs in polynomial time, and outputs a subset S ⊆ V (G) of weight at most d log n · opt(G, w, t)/s such that the treewidth of G−S is at most c·st. Here, d and c are fixed constants that depend only on H, and opt(G, w, t) is the (unknown) minimum weight of a subset U ⊆ V (G) such that the treewidth of G − U is at most t. This lemma immediately yields the first polynomial-time approximation schemes (PTASes) for Weighted Treewidth-η Vertex Deletion, for η ≥ 2, on graphs of bounded genus and the first PTAS for Weighted Feedback vertex Set on H-minor free graphs. These results effortlessly generalize to include weighted edge deletion problems, to all Weighted Connected Planar F-Deletion problems, and finally to quasi polynomial time approximation schemes (QPTASes) for all of these problems on H-minor free graphs. For most of these problems even constant factor approximation algorithms, even on planar graphs, were not previously known. Additionally, using the scaling lemma we subsume, simplify and extend the recent framework of Cohen-Addad et al. [STOC 2016] for turning constant factor approximation algorithms for “ubiquitous” problems into PTASes for the same problems on graphs of bounded genus. Specifically, we obtain PTASes for ubiquitous problems without the requirement of having a constant factor approximation. While the statement of the scaling lemma is inspired by an analogous lemma by Cohen-Addad et al. [STOC 2016] for edge contractions on weighted graphs of bounded genus, as well as a scaling lemma by Fomin et al. [SODA 2011] for unweighted graphs, the proof is entirely different. The proof detours via three different linear programming relaxations for the Weighted Treewidth-η Vertex Deletion problems and a strengthening of a recent rounding procedure of Bansal et al. [SODA 2017] enhanced by the classic Klein-Plotkin-Rao Theorem [STOC 1993].

Original languageEnglish
Title of host publication31st Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2020
EditorsShuchi Chawla
PublisherAssociation for Computing Machinery
Number of pages20
ISBN (Electronic)9781611975994
StatePublished - 1 Jan 2020
Event31st Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2020 - Salt Lake City, United States
Duration: 5 Jan 20208 Jan 2020

Publication series

NameProceedings of the Annual ACM-SIAM Symposium on Discrete Algorithms


Conference31st Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2020
Country/TerritoryUnited States
CitySalt Lake City


Dive into the research topics of 'Approximation schemes via Width/Weight trade-offs on minor-free graphs'. Together they form a unique fingerprint.

Cite this