首页> 外文期刊>Electronic Communications in Probability >How to Combine Fast Heuristic Markov Chain Monte Carlo with Slow Exact Sampling
【24h】

How to Combine Fast Heuristic Markov Chain Monte Carlo with Slow Exact Sampling

机译:如何将快速启发式马尔可夫链蒙特卡罗与慢速精确采样相结合

获取原文
           

摘要

Given a probability law $pi$ on a set $S$ and a function $g : S ightarrow R$, suppose one wants to estimate the mean $ar{g} = int g dpi$. The Markov Chain Monte Carlo method consists of inventing and simulating a Markov chain with stationary distribution $pi$. Typically one has no a priori bounds on the chain's mixing time, so even if simulations suggest rapid mixing one cannot infer rigorous confidence intervals for $ar{g}$. But suppose there is also a separate method which (slowly) gives samples exactly from $pi$. Using $n$ exact samples, one could immediately get a confidence interval of length $O(n^{-1/2})$. But one can do better. Use each exact sample as the initial state of a Markov chain, and run each of these $n$ chains for $m$ steps. We show how to construct confidence intervals which are always valid, and which, if the (unknown) relaxation time of the chain is sufficiently small relative to $m$, have length $O(n^{-1} log n)$ with high probability.
机译:给定一个集合$ S $的概率定律$ pi $和一个函数$ g:S rightarrow R $,假设有人想要估计均值$ bar {g} = int g d pi $。马尔可夫链蒙特卡罗方法包括发明和模拟具有固定分布$ pi $的马尔可夫链。通常情况下,在链的混合时间上没有先验界限,因此即使模拟表明快速混合,也无法推断$ bar {g} $的严格置信区间。但是,假设还有一个单独的方法(缓慢地)准确地给出$ pi $中的样本。使用$ n $个精确样本,可以立即获得长度为$ O(n ^ {-1/2})$的置信区间。但是可以做得更好。将每个精确的样本用作马尔可夫链的初始状态,并以$ m $个步骤运行每个$ n $链。我们展示了如何构造始终有效的置信区间,并且如果链的(未知)弛豫时间相对于$ m / n $足够小,则其长度为$ O(n ^ {-1} log n )$的可能性很高。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号