Mixture decompositions of exponential families using a decomposition of their sample spaces

zbMATH Open1266.60038arXiv1008.0204MaRDI QIDQ4917828FDOQ4917828

Authors: Guido Montúfar 

Publication date: 2 May 2013

Abstract: We study the problem of finding the smallest $m$ such that every element of an exponential family can be written as a mixture of $m$ elements of another exponential family. We propose an approach based on coverings and packings of the face lattice of the corresponding convex support polytopes and results from coding theory. We show that $m=q^{N-1}$ is the smallest number for which any distribution of $N$ $q$-ary variables can be written as mixture of $m$ independent $q$-ary variables. Furthermore, we show that any distribution of $N$ binary variables is a mixture of $m=2^{N-(k+1)}(1+1/(2^k-1))$ elements of the $k$-interaction exponential family.

Full work available at URL: https://arxiv.org/abs/1008.0204

Recommendations

• Exponential family mixture models (with least-squares estimators)
• Local mixture models of exponential families
• On identifiability of mixtures of independent distribution laws
• Optimally approximating exponential families
• Information geometry on hierarchy of probability distributions

zbMATH Keywords

mixture modelperfect codemarginal polytopenon-negative tensor rank

Mathematics Subject Classification ID

Approximations to statistical distributions (nonasymptotic) (62E17) Distribution theory (60E99)

Cites Work

• polymake: a framework for analyzing convex polytopes
• Statistical analysis of finite mixture distributions
• Finite mixture models
• PRMLT
• Mixture models: theory, geometry and applications
• Pattern recognition and machine learning.
• Title not available (Why is that?)
• On the toric algebra of graphical models
• Finite forms of de Finetti's theorem on exchangeability
• Secant varieties of \({\mathbb {P}^1}\times \cdots \times {\mathbb {P}^1} (n\)-times) are not defective for \(n \geq 5\)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• The geometry of exponential families
• Title not available (Why is that?)
• Information geometry on hierarchy of probability distributions
• Maximizing multi-information
• Support sets in exponential families and oriented matroid theory
• Uses of exchangeability
• Title not available (Why is that?)
• Gröbner bases and polyhedral geometry of reducible and cyclic models.
• On the identifiability of binary Segre products
• Prodsimplicial-neighborly polytopes
• Hierarchical Models, Marginal Polytopes, and Linear Codes
• Finding the Maximizers of the Information Divergence From an Exponential Family
• Erratum to: Neighborliness of marginal polytopes
• Refinements of universal approximation results for deep belief networks and restricted Boltzmann machines
• Singular value decomposition of probability matrices: Probabilistic aspects of latent dichotomous variables

Cited In (6)

• Information geometry on hierarchy of probability distributions
• When does a mixture of products contain a product of mixtures?
• Universal Approximation Depth and Errors of Narrow Belief Networks with Discrete Units
• Dimension of Marginals of Kronecker Product Models
• An exact decomposition theorem and a unified view of some related distributions for a class of exponential transformation models on symmetric cones
• Hierarchical models as marginals of hierarchical models

Uses Software

• polymake
• PRMLT

This page was built for publication: Mixture decompositions of exponential families using a decomposition of their sample spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4917828)