Statistical Bergman geometry

Publication date 17 May 2023

Abstract: We introduce an embedding

P h i : O m e g a i g h t a r r o w m a t h c a l P (O m e g a)

, where

O m e g a s u b s e t m a t h b b C^{n}

is a bounded domain, and

m a t h c a l P (O m e g a)

denotes the space of all probability distributions defined on

O m e g a

. We establish that the pullback of the Fisher information metric, the fundamental Riemannian metric in Information geometry, from

m a t h c a l P (O m e g a)

through

P h i

is exactly the Bergman metric of

O m e g a

. By employing the embedding

P h i

, we consider

O m e g a

as a statistical manifold in

m a t h c a l P (O m e g a)

and present several results in this framework. Firstly, we express the holomorphic sectional curvature of the Bergman metric in terms of the expectation and provide a reconfirmation of Kobayashi's result that the holomorphic sectional curvature lies within the range

(- i n f t y, 2]

. Secondly, regarding two bounded domains

O m e g a_{1}

and

O m e g a_{2}

as statistical manifolds, we show that if a proper holomorphic map

f : O m e g a_{1} i g h t a r r o w O m e g a_{2}

preserves the Fisher information metrics, then

f

must be a biholomorphism. Finally, we prove the asymptotic normality with the Bergman metric, subject to mild conditions imposed on the Bergman kernels that are satisfied by bounded Hermitian symmetric domains.

This page was built for publication: Statistical Bergman geometry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6436961)