Examples of the application of nonparametric information geometry to statistical physics

DOI10.3390/E15104042MaRDI QIDQ280599zbMATH OpenOpenAlexFDO

Authors Giovanni Pistone

Publication date 10 May 2016

Published in Entropy (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1308.5312

zbMATH Keywords

Boltzmann equationBoltzmann-Gibbs entropydifferential (statistical) connectionsexponential manifoldinformation geometryKullback-Leibler divergencenon-parametric statistical manifoldsOrlicz spaces

Mathematics Subject Classification ID

Boltzmann equations (35Q20) Spaces of measurable functions ((L^p)-spaces, Orlicz spaces, Köthe function spaces, Lorentz spaces, rearrangement invariant spaces, ideal spaces, etc.) (46E30) Global differential geometry (53C99) Geometric probability and stochastic geometry (60D05) Rarefied gas flows, Boltzmann equation in fluid mechanics (76P05) Kinetic theory of gases in time-dependent statistical mechanics (82C40) Measures of information, entropy (94A17)

Abstract: We review a nonparametric version of Amari's Information Geometry in which the set of positive probability densities on a given sample space is endowed with an atlas of charts to form a differentiable manifold modeled on Orlicz Banach spaces. This nonparametric setting is used to discuss the setting of typical problems in Machine Learning and Statistical Physics, such as relaxed optimization, Kullback-Leibler divergence, Boltzmann entropy, Boltzmann equation

Review text

The departure point of the present review paper is the monograph of \textit{S.-i. Amari} and \textit{H. Nagaoka} [Methods of information geometry. Providence, RI: American Mathematical Society; Oxford: Oxford University Press (2000; Zbl 0960.62005)] where the fundamental treatement of differential connections has been established for finite dimensional coordinate systems. Developments following A. P. Dawid's comments on [\textit{B. Efron}, Ann. Stat. 3, 1189--1242 (1975; Zbl 0321.62013)] enable a non-parametric extension of the Amari-Nagaoka theory where positive probability densities of a measure space were shown to be a differentiable manifold modeled on Orlicz banach spaces. A major purpose of the paper is to show that the pertinent mathematical formalism provides a rigorous way to interpret the Boltzmann equation as an evolution equation on the statistical manifold. Applications of Orlicz spaces in statistical physics have been previously exposed by \textit{W. A. Majewski} and \textit{L. E. Labuschagne} [Ann. Henri Poincaré 15, No. 6, 1197--1221 (2014; Zbl 1295.82019)]. The latter framework is presently worked out for a space-homogeneous Boltzmann operator with a suitable angular collision kernel. The non-parametric setting is also used to discuss a number of problems like e.g statistical physics of machine learning (in particular the black-box optimization), Kullback-Leibler divergence and Boltzmann-Gibbs entropy.

Recommendations

• Nonparametric information geometry
• scientific article; zbMATH DE number 2095988
• Information geometry of the probability simplex: a short course
• Nonparametric information geometry: from divergence function to referential-representational biduality on statistical manifolds
• Extension of Information Geometry to non-statistical systems: some examples

Cites work

• \(\kappa \)-exponential models from the geometrical viewpoint
• A infinite-dimensional geometric structure on the space of all the probability measures equivalent to a given one
• Connections on Non-Parametric Statistical Manifolds by Orlicz Space Geometry
• CONNECTIONS ON STATISTICAL MANIFOLDS OF DENSITY OPERATORS BY GEOMETRY OF NONCOMMUTATIVE L^p-SPACES
• Deformed exponentials and applications to finance
• Dual connections in nonparametric classical information geometry
• Exponential statistical manifold
• Further comments on some comments on a paper by Bradley Efron
• Gradient flows in metric spaces and in the spaces of probability measures, and applications to Fokker-Planck equations with respect to log-concave measures
• scientific article; zbMATH DE number 194094 (Why is no real title available?)
• Information geometry and sufficient statistics
• On -families of probability distributions
• On applications of Orlicz spaces to statistical physics
• Orlicz spaces and modular spaces
• The exponential statistical manifold: mean parameters, orthogonality and space transformations

Cited in

(18)

• The Banach manifold of measures and the Lagrange multipliers of statistical mechanics
• An eikonal equation approach to thermodynamics and the gradient flows in information geometry
• Information geometry of physics-informed statistical manifolds and its use in data assimilation
• Extension of Information Geometry to non-statistical systems: some examples
• New results on mixture and exponential models by Orlicz spaces
• Information geometry formalism for the spatially homogeneous Boltzmann equation
• Exponential models by Orlicz spaces and applications
• A Lecture About the Use of Orlicz Spaces in Information Geometry
• Local measurements of nonlinear embeddings with information geometry
• Lightlike submanifolds of semi-Riemannian statistical manifolds
• Information geometry of the probability simplex: a short course
• Information geometry of smooth densities on the Gaussian space: Poincaré inequalities
• Nonparametric information geometry
• Differential analytic aspects of statistical models
• Affine statistical bundle modeled on a Gaussian Orlicz-Sobolev space
• Lagrangian and Hamiltonian dynamics for probabilities on the statistical bundle
• Sub-exponentiality in statistical exponential models
• Statistical models built on sub-exponential random variables

This page was built for publication: Examples of the application of nonparametric information geometry to statistical physics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q280599)