We present a structural characterization of all tournaments T = (V, A) such that, for any nonnegative integral weight function defined on V , the maximum size of a feedback vertex set packing is equal to the minimum weight of a triangle in T . We also answer a question of Frank by showing that it is NP -complete to decide whether the vertex set of a given tournament can be partitioned into two feedback vertex sets. In addition, we give exact and approximation algorithms for the feedback vertex set packing problem on tournaments.
Introduction.A rich variety of combinatorial optimization problems falls within the general framework of packing and covering in hypergraphs. A hypergraph is a pair H = (V, E), where V is a finite set and E is a family of subsets of V . Elements of V and E are called the vertices and edges of H, respectively. A vertex cover of H is a vertex subset that intersects all edges of H. Let w be a nonnegative integral weight function defined on V . A family S of edges (repetition is allowed) of H is called a w-packing of H if each v ∈ V belongs to at most w(v) members of S. Let ν w (H) denote the maximum size of a w-packing of H, and let τ w (H) denote the minimum total weight of a vertex cover. Clearly ν w (H) ≤ τ w (H); this inequality, however, need not hold equality in general. We say that H is Mengerian if the min-max relation ν w (H) = τ w (H) is satisfied for any nonnegative integral function w defined on V . Many celebrated results and conjectures in combinatorial optimization can be rephrased by saying that certain hypergraphs are Mengerian (see section 79.1 of [19]), so Mengerian hypergraphs have been subjects of extensive research. As conjectured by Edmonds and Giles [9, 18] and proved recently by Ding, Feng, and Zang [4], the problem of recognizing Mengerian hypergraphs is NP -hard in general, and hence it cannot be solved in polynomial time unless NP = P . In this paper we study a special class of Mengerian hypergraphs; our work is a continuation of those done in [1,2,3,5,6].