arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.22237cs.ITmath.ITphysics.soc-ph
收藏

具有结构的发散度用于比较概率分布

Parametric and structure-aware information theory for multi-scale analysis of composition

Mathematical Institute, University of Oxford(牛津大学数学研究所);Institute for New Economic Thinking, University of Oxford(牛津大学新经济思维研究所);Complexity Science Hub Vienna(维也纳复杂科学中心)

Rohit Sahasrabuddhe, Renaud Lambiotte

AI总结:

本文提出了一种敏感于底层领域几何结构的概率分布发散度,用于在合成聚类任务中恢复隐藏模式,并在经济地理和生态学中展示了其应用。

摘要:

许多自然和社会科学系统通过具有相互关系的元素的概率分布来描述:例如,具有共享技能的职业或具有相似特征的物种。标准信息论量如熵和f-发散度将元素视为等同,并对相似性结构无感。我们引入了一种发散度家族,能够感知底层域的几何结构。通过作为结构感知熵的Bregman发散度,它们提供了一个框架,保留了Kullback-Leibler发散度和Shannon熵的多个优点。结构感知发散度在合成聚类任务中恢复隐藏模式,而传统发散度无法做到,并且比最优传输距离快数个数量级。我们展示了其在经济地理和生态学中的应用,其中结构起重要作用。建模不同的职业相关性概念导致其地理分布的定性不同区域化。我们的方法还再现了使用最优传输方法获得的生态学中功能β-多样性已知见解。

英文摘要:

Compositional data is common across the natural and social sciences, requiring methods to measure the diversity of compositions and the heterogeneity of collections of them. Via Bregman geometry, a strictly concave diversity index induces a measure of heterogeneity that decomposes across scales. A canonical example is Shannon entropy inducing mutual information. We develop this framework by incorporating category similarity into $α$-logarithmic entropy, a parametric diversity index. We disprove an existing general concavity result and prove that, for $α=3$, the region of strict concavity is always convex, complementing previous results for $α=1$ and $2$. The resulting geometry provides methods to compare compositions substantially faster than optimal transport. Applied to occupation compositions across England and Wales, they reveal distinct regionalisations based on different notions of occupation relatedness. Applied to ecological data, they recover established patterns of functional and taxonomic $β$-diversity, and reveal sensitivity to the diversity parameter.

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。

↑