A Lower Bound on Cycle-Finding in Sparse Digraphs.
X. Chen and T. Randolph and R. Servedio and T. Sun.
ACM-SIAM Symposium on Discrete Algoithms (SODA), 2020, 2936-2952.


Abstract:

We consider the problem of finding a cycle in a sparse directed graph $G$ that is promised to be far from acyclic, meaning that the smallest feedback arc set in $G$ is large. We prove an information-theoretic lower bound, showing that for $N$-vertex graphs with constant outdegree any algorithm for this problem must make $\tilde{\Omega}(N^{5/9})$ queries to an adjacency list representation of $G$. In the language of property testing, our result is an $\tilde{\Omega}(N^{5/9})$ lower bound on the query complexity of one-sided algorithms for testing whether sparse digraphs with constant outdegree are far from acyclic. This is the first improvement on the $\Omega(\sqrt{N})$ lower bound, implicit in Bender and Ron's 2002 paper, which follows from a simple birthday paradox argument.

pdf of full version


Back to main papers page