Network decomposition is a central concept in the study of distributed graph algorithms. We present the first polylogarithmic-round deterministic distributed algorithm with small messages that constructs a strong-diameter network decomposition with polylogarithmic parameters. Concretely, a ($C$, $D$) strong-diameter network decomposition is a partitioning of the nodes of the graph into disjoint clusters, colored with $C$ colors, such that neighboring clusters have different colors and the subgraph induced by each cluster has a diameter at most $D$. In the weak-diameter variant, the requirement is relaxed by measuring the diameter of each cluster in the original graph, instead of the subgraph induced by the cluster. A recent breakthrough of Rozho\v{n} and Ghaffari [STOC 2020] presented the first $\text{poly}(\log n)$-round deterministic algorithm for constructing a weak-diameter network decomposition where $C$ and $D$ are both in $\text{poly}(\log n)$. Their algorithm uses small $O(\log n)$-bit messages. One can transform their algorithm to a strong-diameter network decomposition algorithm with similar parameters. However, that comes at the expense of requiring unbounded messages. The key remaining qualitative question in the study of network decompositions was whether one can achieve a similar result for strong-diameter network decompositions using small messages. We resolve this question by presenting a novel technique that can transform any black-box weak-diameter network decomposition algorithm to a strong-diameter one, using small messages and with only moderate loss in the parameters.
翻译:网络分解是分布式图表算法研究中的核心概念。 我们展示了第一个多式圆柱形确定式分布式算法, 含有小信息, 构建了强直径网络分解, 并带有多对数参数。 具体地说, 一个( $C, $D$) 强直径网络分解是将图形节点分割成分解组, 彩色为美元, 相邻组的颜色不同, 每个组群的子组的分解值直径最高为$D$。 在弱度直径模型变异变量中, 通过测量原始图中每个集的直径, 而不是集集引出子图。 一个( Rozhov{n} 和 Ghaffari [STOC 2020] 的最近突破是将图形的节点分割成分解组合, 以 $text{poly rom commetage, $C$ and $D$Dgrodegraphle com 两者都以 $m- decom ral requistration com com) exmal deal decomst exmation exmal deal exmation exmation a exmation a exmalmagistration exmation exmationaldia exmalmaxl export a export a exmaltilex ex ex exm exmetaldaldaldaldaldald exm exmationaldaldalizutus.