Computer Science > Information Theory
[Submitted on 12 Apr 2025 (this version), latest version 15 Apr 2025 (v2)]
Title:A Hierarchical Decomposition of Kullback-Leibler Divergence: Disentangling Marginal Mismatches from Statistical Dependencies
View PDF HTML (experimental)Abstract:The Kullback-Leibler (KL) divergence is a foundational measure for comparing probability distributions. Yet in multivariate settings, its structure is often opaque, conflating marginal mismatches and statistical dependencies. We derive an algebraically exact, additive, and hierarchical decomposition of the KL divergence between a joint distribution \( P_k \) and a product reference \( Q^{\otimes k} \). The total divergence splits into the sum of marginal KLs, \( \sum_{i=1}^k \mathrm{KL}(P_i \| Q) \), and the total correlation \( C(P_k) \), which we further decompose as \( C(P_k) = \sum_{r=2}^k I^{(r)}(P_k) \), using Moebius inversion on the subset lattice. Each \( I^{(r)} \) quantifies the distinct contribution of \( r \)-way statistical interactions to the total divergence. This yields the first decomposition of this form that is both algebraically complete and interpretable using only standard Shannon quantities, with no approximations or model assumptions. Numerical validation using hypergeometric sampling confirms exactness to machine precision across diverse system configurations. This framework enables precise diagnosis of divergence origins, marginal versus interaction, across applications in machine learning, econometrics, and complex systems.
Submission history
From: William Cook [view email][v1] Sat, 12 Apr 2025 00:59:07 UTC (949 KB)
[v2] Tue, 15 Apr 2025 21:48:52 UTC (949 KB)
Current browse context:
cs.IT
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.