We give a new rapid mixing result for a natural random walk on the independent sets of an input graph $G$. Rapid mixing is of interest in approximately sampling a structure, over some underlying set or graph, from some target distribution. In the case of independent sets, we show that when $G$ has bounded treewidth, this random walk -- known as the hardcore Glauber dynamics -- mixes rapidly for all values of the standard parameter $\lambda > 0$, giving a simple alternative to existing sampling algorithms for these structures. We also show rapid mixing for Markov chains on dominating sets and $b$-edge covers (for fixed $b\geq 1$ and $\lambda > 0$) in the case where treewidth is bounded, and for Markov chains on the $b$-matchings (for fixed $b \geq 1$ and $\lambda > 0$), the maximal independent sets, and the maximal $b$-matchings of a graph (for fixed $b \geq 1$), in the case where carving width is bounded. We prove our results by developing a divide-and-conquer framework using the well-known multicommodity flows technique. Using this technique, we additionally show that a similar dynamics on the $k$-angulations of a convex set of $n$ points mixes in quasipolynomial time for all $k \geq 3$. (McShine and Tetali gave a stronger result in the special case $k = 3$.) Our technique also allows us to strengthen existing results by Dyer, Goldberg, and Jerrum and by Heinrich for the Glauber dynamics on the $q$-colorings of $G$ on graphs of bounded carving width, when $q \geq \Delta + 2$ is bounded. Specifically, our technique yields an improvement in the dependence on treewidth when $\Delta < 2t$ or when $q < 4t$ and $\Delta < t^2$. We additionally show that the Glauber dynamics on the partial $q$-colorings of $G$ mix rapidly for all $\lambda > 0$ when $q \geq \Delta + 2$ is bounded.
翻译:我们给出了一个新的快速混合结果, 用于在独立输入图集上自然随机行走 $G$。 快速混用对于大约取样一个结构( 超过某些基底的电流集或图形 $$$ 美元 ), 以某些目标分布。 在独立组中, 我们显示当$ 捆绑起来时, 这种随机行走( 被称为硬点的Glauber 动态) 快速混合标准参数的所有值 $\ lambda > 0, 简单替代这些结构的现有采样算法 。 我们还展示了马可夫在支配器和 $ $ 美元 上更坚固的链 。 在使用离子的电磁盘中, 当使用离子的电流中, 3美元 和 美元 的电流中, 我们的电流中, 也显示了更多的电流 。