Nested sampling for physical scientists
面向物理科学家的嵌套采样综述
We review Skilling's nested sampling (NS) algorithm for Bayesian inference and more broadly multi-dimensional integration. After recapitulating the principles of NS, we survey developments in implementing efficient NS algorithms in practice in high-dimensions, including methods for sampling from the so-called constrained prior. We outline the ways in which NS may be applied and describe the application of NS in three scientific fields in which the algorithm has proved to be useful: cosmology, gravitational-wave astronomy, and materials science. We close by making recommendations for best practice when using NS and by summarizing potential limitations and optimizations of NS.
展开 ▾嵌套采样能够同时给出贝叶斯证据和后验估计,无需像退火方法那样设计温度调度;动态嵌套采样大幅提高了参数推断效率,并在多个物理科学领域展现出广泛适用性。
嵌套采样自 Skilling 2004 和 Skilling 2006 提出后,凭借其同时计算证据和后验、处理多模态分布的能力,迅速被宇宙学界采用,例如 Mukherjee+ 2006 和 Feroz+ 2008 引入椭球区域采样,将算法推广到高维多模态问题。随后 Pártay+ 2010 将嵌套采样引入材料科学,用于计算配分函数和相图;在引力波天文学中,Veitch+ 2010 等发展了适用于致密双星信号分析的嵌套采样方法。为克服固定活跃点数的局限,Higson+ 2018 提出动态嵌套采样,允许运行中动态调整活跃点,显著提升了参数推断效率。与序贯蒙特卡洛的联系 Salomone+ 2018 为嵌套采样提供了无偏估计的理论框架,而 Handley+ 2015 的切片采样等步进方法则有效改善了高维约束先验采样的维度可扩展性。展望未来,与机器学习(如正则化流)相结合的嵌套采样有望突破更高维度的限制,并进一步扩展到复杂物理模型的推断与模型选择。
预印本 2022-05-31 · 接收 2022-04-04 · 刊出 2022-05-26 · 收录 2026-09-28