Generalizing the Sharp Threshold Phenomenon for the Distributed Complexity of the Lov\'asz Local Lemma

TL;DR

This paper proves a conjecture that the sharp threshold phenomenon for the distributed Lovász Local Lemma applies in the unrestricted setting, providing an optimal algorithm with tight bounds in bounded-degree graphs.

Contribution

It extends the sharp threshold result to the unrestricted case, introducing a combinatorial approach to prove convexity of certain high-dimensional functions.

Findings

Algorithm solves LLL in $O(d^2 + ext{log}^* n)$ time under the $p2^d < 1$ criterion.

The approach confirms the conjecture for arbitrary dependencies among variables.

Provides tight bounds matching lower bounds in bounded-degree graphs.

Abstract

Recently, Brandt, Maus and Uitto [PODC'19] showed that, in a restricted setting, the dependency of the complexity of the distributed Lov\'asz Local Lemma (LLL) on the chosen LLL criterion exhibits a sharp threshold phenomenon: They proved that, under the LLL criterion $p2^d < 1$, if each random variable affects at most $3$ events, the deterministic complexity of the LLL in the LOCAL model is $O(d^2 + \log^* n)$. In stark contrast, under the criterion $p2^d \leq 1$, there is a randomized lower bound of $\Omega(\log \log n)$ by Brandt et al. [STOC'16] and a deterministic lower bound of $\Omega(\log n)$ by Chang, Kopelowitz and Pettie [FOCS'16]. Brandt, Maus and Uitto conjectured that the same behavior holds for the unrestricted setting where each random variable affects arbitrarily many events. We prove their conjecture, by providing an algorithm that solves the LLL in time $O(d^2 + \log^* n)$ under the LLL criterion $p2^d < 1$, which is tight in bounded-degree graphs due to an $\Omega(\log^* n)$ lower bound by Chung, Pettie and Su [PODC'14]. By the work of Brandt, Maus and Uitto, obtaining such an algorithm can be reduced to proving that all members in a certain family of functions in arbitrarily high dimensions are convex on some specific domain. Unfortunately, an analytical description of these functions is known only for dimension at most $3$, which led to the aforementioned restriction of their result. While obtaining those descriptions for functions of (substantially) higher dimension seems out of the reach of current techniques, we show that their convexity can be inferred by combinatorial means.