REVIEW 4 major objections 3 minor
A Riemannian viewpoint on the Amari-Cencov $\alpha$-connections and Proudman-Johnson equations
T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Every Amari–Cencov α-connection on densities is the Levi-Civita connection of a new metric.
desk verdict A plausible and significant construction that deserves a careful referee; the main open question is whether the weak Riemannian metric has a well-defined Levi-Civita connection in the Fréchet setting. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the family of $\alpha$-Fisher-Rao metrics $G^\alpha$ on $\operatorname{Dens}_+(M)$, constructed so that their Levi-Civita connection coincides with the Amari-Cencov connection $\nabla^{(\alpha)}$. The key is allowing $G^\alpha$ to break $\operatorname{Diff}(M)$-invariance for $\alpha \neq 0$, which gives the metric enough freedom to encode the $\alpha$-dependence while the connection remains invariant. On $\operatorname{Prob}(M)$, the machinery is a radial-projection description of $\nabla^{(\alpha)}$-geodesics, drawing them as radial projections of straight lines onto hypersurfaces, from which geodesic convexity for all real $\alpha$ follows. On $\operatorname{Diff}(M)$, the machinery is the Euler-Arnold formulation of geodesic equations for non-right-invariant metrics, which turns the generalized Proudman-Johnson equations into geodesic equations of such metrics.
What would settle it
Take $M=S^1$, write densities as $\exp(u)$ in a fixed Sobolev space, compute the Levi-Civita connection of $G^\alpha$ directly using the paper's definition, and compare its Christoffel symbols with the standard Amari-Cencov connection for some $\alpha \neq 0$; a single $\alpha$ where the two differ would disprove the central claim.
Extended reading notes
Core claim
The central claim is that the Amari-Cencov $\alpha$-connections $\nabla^{(\alpha)}$ arise as Levi-Civita connections: on $\operatorname{Dens}_+(M)$ there exists a one-parameter family of Riemannian metrics $G^\alpha$—the $\alpha$-Fisher-Rao metrics—such that the Levi-Civita connection of $G^\alpha$ equals $\nabla^{(\alpha)}$. For $\alpha \neq 0$ these metrics are not invariant under the diffeomorphism group $\operatorname{Diff}(M)$, even though the connections are, so the $\alpha$-dependence is absorbed by breaking a symmetry while preserving the affine structure. On $\operatorname{Prob}(M)$ the same phenomenon occurs precisely for $\alpha \in \{-1,0,1\}$, and no metric of this kind exists for other $\alpha$; nevertheless, $\nabla^{(\alpha)}$-geodesics on $\operatorname{Prob}(M)$ can be described as radial projections of straight lines onto suitable hypersurfaces, and this picture gives geodesic convexity for every real $\alpha$. For the diffeomorphism group, analogous non-right-invariant metrics have the generalized Proudman-Johnson equations on the real line as their Euler-Arnold equations. In finite-dimensional statistical models, whether $\nabla^{(\alpha)}$ is metric depends on the model, not just on $\alpha$.
Load-bearing premise
The construction requires that the space of densities can carry these new metrics without requiring them to be unchanged under all smooth reparametrizations; if reparametrization invariance is forced, the result collapses to the $\alpha=0$ case.
Editorial extensions
If this is right
- For every $\alpha$, $\nabla^{(\alpha)}$-geodesics on $\operatorname{Dens}_+(M)$ can be viewed as energy-minimizing curves for the $\alpha$-Fisher-Rao metric, making variational tools (existence, uniqueness, convexity) applicable to $\alpha$-geodesics.
- On $\operatorname{Prob}(M)$, the $\alpha$-connections are Levi-Civita connections only for $\alpha = -1, 0, 1$; for other $\alpha$ they are not metric, yet the radial-projection description still yields geodesic convexity for every real $\alpha$.
- The generalized Proudman-Johnson equations on the real line are the Euler-Arnold equations of non-right-invariant metrics on $\operatorname{Diff}(M)$, giving them a geodesic interpretation and an associated energy functional.
- Finite-dimensional statistical models inherit metricity of $\nabla^{(\alpha)}$ in a model-dependent way, so the question 'is an $\alpha$-connection metric?' has no $\alpha$-only answer.
Reading between the lines
- An implicit consequence is that $\operatorname{Dens}_+(M)$ carries many Riemannian structures with the same Levi-Civita connection but different geodesic energies, so choosing an energy functional for $\alpha$-geodesics involves a freedom not visible from the connection alone.
- The non-invariance of $G^\alpha$ hints at a link to optimal-transport and Wasserstein-type geometries, which are also non-invariant and exhibit geodesic-convexity phenomena; probing whether $\nabla^{(\alpha)}$-geodesics relate to Wasserstein geodesics for specific $\alpha$ would be a natural extension.
- For $M = S^1$, the same Euler-Arnold machinery with non-right-invariant metrics may produce integrable partial differential equations beyond the classical right-invariant cases (Camassa-Holm, Hunter-Saxton), and computing their Hamiltonians explicitly would be a testable extension.
- The radial-projection construction on $\operatorname{Prob}(M)$ may yield explicit formulas for the exponential map and Jacobi fields of $\nabla^{(\alpha)}$, since straight lines on the ambient hypersurface have known Jacobi behavior; this could give curvature comparisons for all $\alpha$.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims a new geometric interpretation of the Amari–Cencov alpha-connections in information geometry. On the space of positive densities Dens_+(M), it asserts the existence of one-parameter families of Riemannian metrics G^α, called alpha-Fisher–Rao metrics, whose Levi-Civita connections coincide with ∇^(α). For α ≠ 0 these metrics are not invariant under the diffeomorphism group, even though the connections are invariant. On the space of probability densities, the paper reports that the same metric description exists only for α ∈ {-1, 0, 1} and that α-connections are not metric otherwise. It further claims that ∇^(α)-geodesics on Prob(M) can be described as radial projections of straight lines on hypersurfaces, yielding geodesic convexity for all real α. Additional results are announced for Diff(M), relating generalized Proudman–Johnson equations to Euler–Arnold equations of non-right-invariant metrics, and for finite-dimensional statistical models, where metricity depends on the model.
Significance. If the announced results are correct, the paper provides a unified Riemannian-geometric interpretation of all Amari–Cencov alpha-connections, giving geodesics an energy-minimizing characterization. The explicit construction of non-Diff-invariant metrics with prescribed Levi-Civita connections is conceptually striking and could open new links between information geometry, infinite-dimensional Riemannian geometry, and hydrodynamics. The paper is careful in stating exceptions (Prob(M) restricted to α ∈ {-1,0,1}; finite-dimensional dependence on the model), which suggests a mathematically mature treatment. The potential implications for the Proudman–Johnson equations are also notable. However, the significance rests on the correctness of the infinite-dimensional geometric constructions, which are delicate.
major comments (4)
- [Abstract, opening claim] The central existence theorem — that on Dens_+(M) there exist Riemannian metrics G^α whose Levi-Civita connections equal ∇^(α) — is stated without specifying the infinite-dimensional manifold structure or the regularity of the metrics. On a Fréchet manifold with a weak Riemannian metric, the Levi-Civita connection is not guaranteed to exist or be unique; the paper must either construct the connection explicitly and prove metric compatibility and torsion-freeness, or state a precise functional-analytic framework (e.g., ILH or Sobolev completions) in which the Koszul formula is justified. This point is load-bearing because the entire paper rests on this existence result.
- [Abstract, Prob(M) classification] The classification on Prob(M) — metric if and only if α ∈ {-1,0,1} — is a sharp dichotomy. The abstract gives no indication of the proof, and the non-metricity direction is not a formal consequence of the Dens_+(M) result, because passing to the submanifold of probability densities may destroy metricity even if the ambient metric exists. A rigorous proof of the non-existence for all other α must be supplied.
- [Abstract, geodesic convexity] The claim of geodesic convexity for every α ∈ ℝ is surprising in view of the non-metricity for most α on Prob(M). Since geodesics of a non-metric connection are not minimizers of any obvious energy, the proof of convexity must carefully define the geodesic equation and the convexity functional; the radial-projection picture should be made precise, including its domain and the relevant hypersurfaces.
- [Abstract, Diff(M) and Proudman–Johnson equations] The statement that generalized Proudman–Johnson equations are Euler–Arnold equations of non-right-invariant metrics is strong, especially for M = ℝ, a non-compact manifold. The paper should clarify the function space (e.g., Sobolev or smooth diffeomorphisms), the precise definition of the metric, and the sense in which the Euler–Arnold equation is defined for a non-invariant metric, including well-posedness or a formal geometric derivation.
minor comments (3)
- [Abstract, notation] The family of metrics G^α would be clearer if the abstract or the introduction explicitly displayed the dependence on the density f, since the non-invariance comes from a nontrivial weight that is not visible in the notation.
- [Abstract, wording] The phrase 'appropriate hyper-surfaces' is vague; naming the hypersurfaces (e.g., level sets of the normalization functional) would improve the readability and make the claim more checkable.
- [General typography] The name 'Amari-Cencov' should be written as 'Amari–Cencov' with an en-dash, and the accent on 'Cencov' should be consistent throughout the manuscript.
Circularity Check
No circularity: the α-Fisher-Rao metric construction is an explicit existence theorem, not a fitted or self-referential definition.
full rationale
The central claim of the abstract is an existence theorem: on Dens_+(M), there exist Riemannian metrics G^α whose Levi-Civita connections are the Amari–Cencov α-connections ∇^(α). On the evidence available, this is a constructive mathematical result that can be verified by direct computation: the natural one-parameter family of weighted L² metrics G^α_f(h,k)=∫ h k f^{β(α)} dμ has a Levi-Civita connection whose Christoffel term is proportional to hk/f, and the coefficient can be matched to the α-connection by an affine choice of β(α). The metric is not defined as 'the metric whose Levi-Civita connection equals ∇^(α)' in the abstract; instead the abstract asserts existence and introduces a name for the metrics. There is no fitting of parameters to data, no prediction of a quantity that was used as an input, no reliance on self-citations, and no imported uniqueness theorem. The non-invariance under Diff(M) for α≠0 is an enabling feature of the construction rather than a circular premise. The remaining functional-analytic question — whether the weak Riemannian metric on the Fréchet manifold Dens_+(M) admits a unique Levi-Civita connection — is a substantive correctness risk, not a circularity. The abstract alone gives no indication that the derivation reduces to its own inputs by definition. Score 0.
Assumptions & free parameters
assumptions (3)
- domain assumption The space Dens_+(M) is a smooth infinite-dimensional manifold with well-defined tangent spaces, connections, and Levi-Civita metrics.
- domain assumption The Amari-Cencov alpha-connections are defined as in the information geometry literature, including on infinite-dimensional density spaces.
- domain assumption The Euler-Arnold formalism applies to the diffeomorphism group Diff(M) with non-right-invariant metrics.
Cite this review
Pith. "Pith review of A Riemannian viewpoint on the Amari-Cencov $\alpha$-connections and Proudman-Johnson equations." pith.science (2026). https://pith.science/paper/LV3N6RV4
@misc{pith2026250800371,
author = {Pith},
title = {Pith review of: A Riemannian viewpoint on the Amari-Cencov $\alpha$-connections and Proudman-Johnson equations},
year = {2026},
howpublished = {\url{https://pith.science/paper/LV3N6RV4}},
note = {Machine review of arXiv:2508.00371}
}
abstract
We give a new geometric interpretation of the Amari-Cencov $\alpha$-connections $\nabla^{(\alpha)}$ from information geometry: On the space of densities $\operatorname{Dens}_+(M)$, we show that there exist Riemannian metrics $G^\alpha$, which we call $\alpha$-Fisher-Rao metrics, whose Levi-Civita connections are $\nabla^{(\alpha)}$. With the exception of $\alpha=0$ (the Fisher-Rao metric), these metrics are non-invariant to the action of the diffeomorphism group $\operatorname{Diff}(M)$, even though the connections are invariant. This gives a new way of interpreting the geodesics of the $\nabla^{(\alpha)}$ as energy-minimizing curves. On the space of probability densities $\operatorname{Prob}(M)$, we show that the same phenomenon holds for $\alpha\in \{-1,0,1\}$ and that the $\alpha$-connections are not metric otherwise. We show that $\nabla^{(\alpha)}$-geodesics on this space can be interpreted as radial projections of straight lines on appropriate hyper-surfaces, and use this geometric picture to obtain geodesic convexity for any $\alpha\in \mathbb{R}$. In addition, we prove analogous results for appropriate metrics and connections on $\operatorname{Diff}(M)$, which, for the case $M=\mathbb{R}$, imply that the generalized Proudman-Johnson equations on the real line are the Euler-Arnold equations of non-right invariant metrics. Finally, in the finite-dimensional case, we show that $\nabla^{(\alpha)}$ can be metric or non-metric depending on the considered statistical model.
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.