Linear Time Parameterized Algorithms for S ubset F eedback V ertex S et

Abstract

In the S ubset F eedback V ertex S et (S ubset FVS) problem, the input is a graph G on n vertices and m edges, a subset of vertices T , referred to as terminals, and an integer k . The objective is to determine whether there exists a set of at most k vertices intersecting every cycle that contains a terminal. The study of parameterized algorithms for this generalization of the F eedback V ertex S et problem has received significant attention over the past few years. In fact, the parameterized complexity of this problem was open until 2011, when two groups independently showed that the problem is fixed parameter tractable. Using tools from graph minors,, Kawarabayashi and Kobayashi obtained an algorithm for S ubset FVS running in time O ( f ( k )ċ n 2 m ) [SODA 2012, JCTB 2012]. Independently, Cygan et al. [ICALP 2011, SIDMA 2013] designed an algorithm for S ubset FVS running in time 2 O ( k log k ) ċ n O (1) . More recently, Wahlström obtained the first single exponential time algorithm for S ubset FVS, running in time 4 k ċ n O (1) [SODA 2014]. While the 2 O ( k ) dependence on the parameter k is optimal under the Exponential Time Hypothesis, the dependence of this algorithm as well as those preceding it, on the input size is at least quadratic. In this article, we design the first linear time parameterized algorithms for S ubset FVS. More precisely, we obtain the following new algorithms for S ubset FVS. — A randomized algorithm for S ubset FVS running in time O (25.6 k ċ ( n + m )). — A deterministic algorithm for S ubset FVS running in time 2 O ( k log k ) ċ ( n + m ). Since it is known that assuming the Exponential Time Hypothesis, S ubset FVS cannot have an algorithm running in time 2 o ( k ) n O (1) , our first algorithm obtains the best possible asymptotic dependence on both the parameter as well as the input size. Both of our algorithms are based on “cut centrality,” in the sense that solution vertices are likely to show up in minimum size cuts between vertices sampled from carefully chosen distributions.