Matroid Connectedness and Component Decomposition: Breaking the Quadratic Barrier
Manuel Aprile, Roberto Ronco
原文摘要
Deciding whether a matroid is connected and computing its connected components are two fundamental algorithmic problems in matroid theory. Since the 1970s, the best known algorithms for both problems have used $O(n^2)$ independence queries on an $n$-element ground set. Our main contribution is to break this quadratic barrier: we give a deterministic algorithm that computes all connected components of a matroid on $n$ elements with $O(n \log{n})$ independence queries. Our algorithm for component decomposition combines fundamental basis exchanges with a disjoint-set union data structure, avoiding the explicit construction of the fundamental graph. Moreover, we show that our algorithm is asymptotically optimal by giving a lower bound based on a simple information-theoretic argument: any (deterministic or bounded-error randomized) algorithm to compute connected components needs $Ω(n \log{n})$ independence queries. Then, we turn to the query complexity of deciding whether a matroid is connected. For this problem, we present an algorithm that uses $O(nh)$ independence queries, where $h \ge 1$ bounds the size of any chain of nonempty proper cyclic flats. The algorithm achieves asymptotically optimal linear query complexity on elementary split matroids, that include all paving matroids, and on arbitrary direct sums of uniform matroids.
cs.DM