Distributed Testing of Conductance

Abstract: We study the problem of testing conductance in the setting of distributed computing and give a two-sided tester that takes $\mathcal{O}(\log(n) / (\epsilon \Phi^2))$ rounds to decide if a graph has conductance at least $\Phi$ or is $\epsilon$-far from having conductance at least $\Phi^2 / 1000$ in the distributed CONGEST model. We also show that $\Omega(\log n)$ rounds are necessary for testing conductance even in the LOCAL model. In the case of a connected graph, we show that we can perform the test even when the number of vertices in the graph is not known a priori. This is the first two-sided tester in the distributed model we are aware of. A key observation is that one can perform a polynomial number of random walks from a small set of vertices if it is sufficient to track only some small statistics of the walks. This greatly reduces the congestion on the edges compared to tracking each walk individually.