Information Geometry via the Q-Root Transform
Levin Maier

TL;DR
This paper develops an infinite-dimensional ll^p-information geometry framework using the q-root transform, enabling explicit solutions to gradient flows, connections to Hamiltonian systems, and extensions to noncompact manifolds.
Contribution
It introduces a novel ll^p-geometry framework with the q-root transform, constructing Fisher--Rao geometries and analyzing gradient flows in infinite dimensions.
Findings
Explicit solutions for gradient flows under the ll^2-Fisher--Rao metric
Demonstration of geodesic completeness of the e-connection
Connection of flows to a completely integrable Hamiltonian system
Abstract
In this paper, we introduce \emph{-information geometry}, an infinite-dimensional framework that shares key features with the geometry of the space of probability densities \( \mathrm{Dens}(M) \) on a closed manifold, while also incorporating aspects of measure-valued information geometry. We define the \emph{-probability simplex} with a noncanonical differentiable structure induced via the \emph{-root transform} from an open subset of the \( \ell^q \)-sphere. This choice makes the \(q\)-root transform an \emph{isometry} and allows us to construct the \(\ell^2\)- and \(\ell^q\)-Fisher--Rao geometries, including \emph{Amari--\v{C}encov \(\alpha\)-connections} and a \emph{Chern connection} in the \(\ell^q\)-setting. We then apply this framework to an infinite-dimensional linear optimization problem. We show that the corresponding gradient flow with respect to the…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsStatistical Mechanics and Entropy · Topological and Geometric Data Analysis · Stochastic Gradient Optimization Techniques
