Nested sampling (NS) computes parameter posterior distributions and makes Bayesian model comparison computationally feasible. Its strengths are the unsupervised navigation of complex, potentially multi-modal posteriors until a well-defined termination point. A systematic literature review of nested sampling algorithms and variants is presented. We focus on complete algorithms, including solutions to likelihood-restricted prior sampling, parallelisation, termination and diagnostics. The relation between number of live points, dimensionality and computational cost is studied for two complete algorithms. A new formulation of NS is presented, which casts the parameter space exploration as a search on a tree data structure. Previously published ways of obtaining robust error estimates and dynamic variations of the number of live points are presented as special cases of this formulation. A new online diagnostic test is presented based on previous insertion rank order work. The survey of nested sampling methods concludes with outlooks for future research.
翻译:嵌套抽样(nested sampling, NS)可计算参数后验分布并使贝叶斯模型比较成为可能。其优点是无监督地导航复杂的、潜在多峰的后验分布,直至达到定义良好的终止点。本文对嵌套抽样算法和其变体进行了系统的文献综述,重点关注其完整算法,包括解决似然减缩先验抽样、并行化、终止和诊断的方法。本文评估了两种完整算法的活动点数、维度和计算成本之间的关系。本文还提出了一种新的NS公式,将参数空间探索视为对树形数据结构的搜索。先前发表的获取可靠误差估计和动态改变活动点数的方法也可以看作这种方法的特例。本文还提出了一项基于先前插入排名的在线诊断测试。最后,本文展望了未来嵌套抽样方法的研究方向。