In this paper we investigate how graph problems that are NP-hard in general, but polynomially solvable on split graphs, behave on input graphs that are close to being split. For this purpose we define split+ke and split+kv graphs to be the graphs that can be made split by removing at most k edges and at most k vertices, respectively. We show that problems like treewidth and minimum fill-in are fixed parameter tractable with parameter k on split+fee graphs. Along with positive results of fixed parameter tractability of several problems on split+ke and split+kv graphs, we also show a surprising hardness result. We prove that computing the minimum fill-in of split+kv graphs is NP-hard even for k = 1. This implies that also minimum fill-in of chordal+kv graphs is NP-hard for every k. In contrast, we show that the treewidth of split+1v graphs can be computed in polynomial time. This gives probably the first graph class for which the treewidth and the minimum fill-in problems have different computational complexity.
展开▼
