The $Aldous\text{-}Broder$ and $Wilson$ are two well-known algorithms to generate uniform spanning trees (USTs) based on random walks. This work studies their relationship while they construct random trees with the goal of reducing the total time required to build the spanning tree. Using the notion of $branches$ $-$ paths generated by the two algorithms on particular stopping times, we show that the trees built by the two algorithms when running on a complete graph are statistically equivalent on these stopping times. This leads to a hybrid algorithm that can generate uniform spanning trees of complete graphs faster than either of the two algorithms. An efficient two-stage framework is also proposed to explore this hybrid approach beyond complete graphs, showing its feasibility in various examples, including transitive graphs where it requires 25% less time than $Wilson$ to generate a UST.
翻译:$Aldous\ text{-}Broder$ 和 $Wilson$ 是两个众所周知的算法, 用来在随机行走的基础上产生统一的横贯树( USTs) 。 这项工作研究它们之间的关系, 同时它们建造随机树, 目的是减少建造横贯树所需的全部时间。 我们使用两种算法在特定停停时产生的 $branches $-$ 路径的概念, 我们显示, 两种算法在完整图表上运行时所建造的树在统计上与这些停用时间相当。 这导致一种混合算法, 它可以产生比两种算法中任何一个都快的整条图( USTs) 的统一横贯树。 还提出一个高效的两阶段框架, 在完整图表之外探索这种混合方法, 并在多个例子中显示其可行性, 包括中转图, 它需要比美元少25%的时间来生成一个UST。