Uniformity Testing for the Monomer-Dimer Model

Abstract

We study uniformity testing for the monomer-dimer model on a known graph $G=(V,E)$. Given samples from an unknown distribution $D_w(M)\propto \prod_{e\in M} w_e$ over matchings $M$ of $G$, we ask whether $D_w$ is the uniform distribution over all matchings of $G$ or is $\varepsilon$-far from it in total variation distance. Our main result is that, for every maximum-degree bound $Δ$, uniformity can be tested with $\widetilde O_Δ(\sqrt m/\varepsilon^2)$ samples, where $m=|E|$. We complement this with an information-theoretic lower bound for general graphs: for every pair of parameters $n\le m\le n^2$, there are graphs with $Θ(n)$ vertices and $Θ(m)$ edges on which every tester requires $Ω(\sqrt m/\varepsilon^2)$ samples. Thus, for bounded-degree graph families with $m=Θ(n)$, testing uniformity over the space of matchings has $\widetilde Θ(\sqrt n/\varepsilon^2)$ sample complexity, in contrast with the $Θ(n/\varepsilon^2)$ samples needed to estimate $D_w$ in total variation distance.