arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2305.03560math.PRmath.STstat.TH
收藏

对祖先过程转移概率公式的反例

Counterexample to a transition probability formula for the ancestral process

Sylvain Rubenthaler

AI总结:

针对描述加权粒子系统时间反转后诱导的划分值过程转移概率的知名公式,本文构造反例予以反驳,该系统涵盖广泛应用于应用统计的序贯蒙特卡洛方法。

摘要:

我们研究加权粒子系统,其中新一代粒子从当前粒子中按与权重成正比的概率重采样得到。这类系统涵盖了广泛应用于应用统计学中的一大类序贯蒙特卡洛(Sequential Monte Carlo)方法。我们考察嵌入此类粒子系统中的系谱树。当时间反转时,该粒子系统会诱导出一族取值为划分的过程(即系谱树叶节点上的划分)。本文的目标是给出一个反例,用以反驳描述该过程转移概率的一个知名公式。

英文摘要:

We consider weighted particle systems in which new generations are re-sampled from current particles with probabilities proportional to their weights. This covers a broad class of sequential Monte Carlo methods, widely used in applied statistics. We consider the genealogical tree embedded into such particle system. When the time is reversed, the particle system induces a partition valued family of processes (partitions on the leaves of the genealogical tree). Our aim here is to give a counterexample to a well known formula describing the transition probabilities of this process.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑