Infinite-dimensional statistical manifolds based on a balanced chart

Foundations and philosophical topics in statistics (62A01) Geometric probability and stochastic geometry (60D05) Probability theory on algebraic and topological structures (60B99)

Abstract: We develop a family of infinite-dimensional Banach manifolds of measures on an abstract measurable space, employing charts that are "balanced" between the density and log-density functions. The manifolds,

(i l d e M_{l a m b d a}, l a m b d a i n [2, i n f t y))

, retain many of the features of finite-dimensional information geometry; in particular, the

a l p h a

-divergences are of class

C^{l c e i l l a m b d a c e i l - 1}

, enabling the definition of the Fisher metric and

a l p h a

-derivatives of particular classes of vector fields. Manifolds of probability measures,

(M_{l a m b d a}, l a m b d a i n [2, i n f t y))

, based on centred versions of the charts are shown to be

C^{l c e i l l a m b d a c e i l - 1}

-embedded submanifolds of the

i l d e M_{l a m b d a}

. The Fisher metric is a pseudo-Riemannian metric on

i l d e M_{l a m b d a}

. However, when restricted to finite-dimensional embedded submanifolds it becomes a Riemannian metric, allowing the full development of the geometry of

a l p h a

-covariant derivatives.

i l d e M_{l a m b d a}

and

M_{l a m b d a}

provide natural settings for the study and comparison of approximations to posterior distributions in problems of Bayesian estimation.

Recommendations

Cites work

Cited in

(9)

This page was built for publication: Infinite-dimensional statistical manifolds based on a balanced chart

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q265266)