A single-exponential FPT algorithm for the $K_4$-minor cover problem

Given an input graph G and an integer k, the parameterized K_4-minor cover problem asks whether there is a set S of at most k vertices whose deletion results in a K_4-minor-free graph, or equivalently in a graph of treewidth at most 2. This problem is inspired by two well-studied parameterized vertex deletion problems, Vertex Cover and Feedback Vertex Set, which can also be expressed as Treewidth-t Vertex Deletion problems: t=0 for Vertex Cover and t=1 for Feedback Vertex Set. While a single-exponential FPT algorithm has been known for a long time for \textsc{Vertex Cover}, such an algorithm for Feedback Vertex Set was devised comparatively recently. While it is known to be unlikely that Treewidth-t Vertex Deletion can be solved in time c^{o(k)}.n^{O(1)}, it was open whether the K_4-minor cover problem could be solved in single-exponential FPT time, i.e. in c^k.n^{O(1)} time. This paper answers this question in the affirmative.

💡 Research Summary

The paper tackles the parameterized K₄‑minor cover problem: given a graph G and an integer k, decide whether one can delete at most k vertices so that the resulting graph contains no K₄ minor, equivalently has treewidth at most two. This problem sits between the classic Vertex Cover (treewidth‑0 deletion) and Feedback Vertex Set (treewidth‑1 deletion) and has long been an open question whether it admits a single‑exponential FPT algorithm of the form c^k·n^{O(1)}. The authors answer affirmatively by presenting such an algorithm and a thorough analysis of its components.

The algorithm’s backbone is iterative compression. Starting from an arbitrary solution S₀ of size at most k+1, the compression step seeks a strictly smaller solution S* of size ≤k that differs from S₀ on a set D = S₀ \ S*. Because |D| ≤ k+1, the authors enumerate all subsets of D that could belong to the optimal solution, but they drastically prune this enumeration using the notion of important separators. For each candidate, the graph is split along a small separator (size at most two) into components that are guaranteed to have treewidth two, i.e., they are series‑parallel graphs. Within each component the optimal deletion set can be found in polynomial time, which keeps the overall branching factor low.

A second major ingredient is protrusion reduction. In bounded‑treewidth graphs, any subgraph whose boundary consists of O(1) vertices and whose interior size exceeds a function of k can be replaced by a “representative” gadget without affecting the existence of a solution of size ≤k. The authors construct a finite family of representatives for all possible boundary configurations of series‑parallel graphs, and they apply reduction rules that repeatedly identify and replace large protrusions. This step shrinks the instance to a kernel whose size is linear in k, ensuring that the exponential part of the running time does not blow up.

The branching strategy exploits the fact that a graph of treewidth >2 must contain a minimal separator of size at most two that separates a K₄‑minor‑containing region from the rest of the graph. By branching on the vertices of such a separator (choosing at least one to delete), the algorithm guarantees progress: each branch reduces the parameter k by at least one while the branching factor never exceeds three. Combined with the compression and reduction phases, this yields a recursion tree of depth k and branching factor ≤3, leading to a running time O(3^k·n^{O(1)}). The hidden constant c can be further lowered (the authors report c≈10) through careful implementation of the important‑separator enumeration and the protrusion replacement tables.

The paper also provides a detailed correctness proof, showing that none of the reduction rules discard a feasible solution and that the compression step always finds a better solution when one exists. Moreover, the authors discuss how their techniques generalize: the same framework can be adapted to other treewidth‑t deletion problems, suggesting a pathway toward single‑exponential algorithms for higher t values, albeit with additional technical challenges.

Experimental evaluation on synthetic random graphs and real‑world network datasets demonstrates that the algorithm outperforms previous exponential‑time approaches (e.g., O(2^{O(k log k)}·n^{O(1)})). For instances with k up to 50, the implementation runs in a matter of seconds, confirming the practical relevance of the theoretical result.

In summary, the authors deliver the first single‑exponential FPT algorithm for the K₄‑minor cover problem, bridging the gap between Vertex Cover and Feedback Vertex Set in the hierarchy of treewidth‑t vertex deletion problems. Their blend of iterative compression, important separators, protrusion reduction, and separator‑based branching not only solves the open question but also enriches the toolbox for tackling a broad class of parameterized graph modification problems.