Nonparametric information geometry: from divergence function to referential-representational biduality on statistical manifolds (Q280743): Difference between revisions

Summary: Divergence functions are the non-symmetric ``distance'' on the manifold \(\mathcal M_\theta\) of parametric probability density functions over a measure space \((X,\mu)\). Classical information geometry prescribes on \(\mathcal M_\theta\): (i) a Riemannian metric given by the Fisher information; (ii) a pair of dual connections (giving rise to the family of \(\alpha\)-connections) that preserve the metric under parallel transport by their joint actions; and (iii) a family of divergence functions \(\alpha\)-divergence) defined on \(\mathcal M_\theta\times\mathcal M_\theta\), which induce the metric and the dual connections. Here, we construct an extension of this differential geometric structure from \(\mathcal M_\theta\) (that of parametric probability density functions) to the manifold \(\mathcal M\) of non-parametric functions on \(X\), removing the positivity and normalization constraints. The generalized Fisher information and \(\alpha\)-connections on \(\mathcal M\) are induced by an \(\alpha\)-parameterized family of divergence functions, reflecting the fundamental convex inequality associated with any smooth and strictly convex function. The infinite-dimensional manifold \(\mathcal M\) has zero curvature for all these \(\alpha\)-connections; hence, the generally non-zero curvature of \(\mathcal M_\theta\) can be interpreted as arising from an embedding of \(\mathcal M_\theta\) into \(\mathcal M\). Furthermore, when a parametric model (after a monotonic scaling) forms an affine submanifold, its natural and expectation parameters form biorthogonal coordinates, and such a submanifold is dually flat for \(\alpha=\pm 1\), generalizing the results of Amari's \(\alpha\)-embedding. The present analysis illuminates two different types of duality in information geometry, one concerning the referential status of a point (measurable function) expressed in the divergence function (``referential duality'') and the other concerning its representation under an arbitrary monotone scaling (``representational duality'').