arxiv: 2603.25396 · v2 · submitted 2026-03-26 · 🧮 math.OC · math.DG

Recognition: 2 theorem links

· Lean Theorem

Optimization on Weak Riemannian Manifolds

Valentina Zalbertus , Max Pfeffer , Alexander Schmeding

Authors on Pith no claims yet

Pith reviewed 2026-05-15 00:33 UTC · model grok-4.3

classification 🧮 math.OC math.DG

keywords weak Riemannian manifoldgradient descentHesse manifoldshape analysisshape optimizationinfinite-dimensional optimizationRiemannian geometrygradient flow

0 comments

The pith

Weak Riemannian manifolds support gradient descent optimization through the introduction of a Hesse manifold structure.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a framework for gradient descent on weak Riemannian manifolds that arise in shape analysis, where the underlying spaces are infinite-dimensional and not modeled on Banach spaces. It defines a Hesse manifold to ensure descent directions and gradient flows remain well-defined under weak metrics. This approach matters because standard strong Riemannian assumptions often fail in practical shape optimization, limiting direct application of descent methods. The work then verifies basic convergence and existence properties for several concrete classes of manifolds used in shape analysis.

Core claim

Optimization via gradient descent extends to weak Riemannian manifolds by defining a Hesse manifold that carries a well-defined gradient and descent direction; foundational properties such as existence of minimizing sequences and first-order optimality conditions then hold for classes of these manifolds arising in shape analysis and shape optimization.

What carries the argument

The Hesse manifold, a weak Riemannian manifold equipped with a structure that guarantees a gradient operator and descent direction for smooth objective functions despite the absence of a strong metric.

If this is right

Shape optimization problems on infinite-dimensional manifolds can be attacked directly with gradient descent without first strengthening the metric.
First-order necessary conditions for optimality become available on the target classes of weak Riemannian manifolds.
Gradient flows exist and can be used to compute critical points on several manifolds arising in shape analysis.
The framework supplies a uniform language for comparing descent algorithms across different weak structures.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same construction may apply to other weak geometric structures outside shape analysis, such as certain spaces of mappings or measures.
Numerical implementations could test whether the theoretical descent directions remain stable under discretization of the underlying manifold.
If the Hesse manifold condition holds, one could derive second-order methods by adding a Hessian-like operator consistent with the weak metric.

Load-bearing premise

A weak Riemannian metric on an infinite-dimensional manifold is enough to define a gradient and a descent direction for optimization without extra regularity that would turn the setting into a strong Riemannian or Banach one.

What would settle it

A concrete example of a weak Riemannian manifold and a smooth function on it for which no descent direction exists at a non-critical point would show the framework does not hold.

Figures

Figures reproduced from arXiv: 2603.25396 by Alexander Schmeding, Max Pfeffer, Valentina Zalbertus.

**Figure 1.** Figure 1: Riemannian gradient descent for f. Left: evolution of the iterates. Right: function values and gradient norms over twenty iterations [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗

**Figure 2.** Figure 2: illustrates the behavior of the Riemannian gradient descent with constant step-size α “ 0.04 and parameter λ “ 0.7. The left panel shows the evolution of the iterates ck under the Riemannian gradient descent. The right panel depicts the decrease of the function value fgpckq´fgpcminq in norm, together with the norm of the Riemannian gradient }∇fgpckq}ck , over twenty iterations [PITH_FULL_IMAGE:figures/ful… view at source ↗

read the original abstract

Riemannian structures on infinite-dimensional manifolds arise naturally in shape analysis and shape optimization. These applications lead to optimization problems on manifolds which are not modeled on Banach spaces. The present article develops the basic framework for optimization via gradient descent on weak Riemannian manifolds leading to the notion of a Hesse manifold. Further, foundational properties for optimization are established for several classes of weak Riemannian manifolds connected to shape analysis and shape optimization.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper introduces Hesse manifolds to set up gradient descent on weak Riemannian manifolds for shape analysis, but the gradient construction may not stay inside the tangent bundle without extra regularity that the weak metric does not automatically supply.

read the letter

The main point is that this work tries to build a basic gradient-descent framework for manifolds equipped with weak Riemannian metrics, which arise in infinite-dimensional shape spaces that are not Banach. They define a Hesse manifold as the setting where the necessary descent properties hold and then record some foundational facts for a few concrete classes tied to shape optimization. That extension is the actual new piece; standard Riemannian optimization results stop short of these weaker structures, so the motivation is reasonable and the connection to shape analysis is direct. They do a clean job stating why the usual Banach or strong-metric assumptions fail here and why a tailored notion is needed. The definitions and claimed properties look formally consistent on their own terms. The soft spot is exactly the one flagged in the stress-test note. For the gradient to be an intrinsic vector field on the manifold, the weak metric must still give a continuous isomorphism from the cotangent to the tangent bundle. In the infinite-dimensional examples they cite, weak metrics often produce only distributional objects unless coercivity or completion is added, which would collapse the setting back to a stronger one. The paper proceeds with the formal setup, but if the isomorphism is not proved or checked against the shape-space cases, the descent flow is not guaranteed to be well-defined on the original manifold. That is a load-bearing gap rather than a minor omission. The paper is aimed at people already working on infinite-dimensional geometry and shape optimization who need a reference for the weak-metric case. A reader looking for theoretical groundwork rather than ready algorithms could extract the definitions and basic properties. It is not yet at the stage of solving open problems or delivering new algorithms, but the framing is honest. I would send it to peer review so that referees can verify whether the gradient map is actually constructed or merely assumed.

Referee Report

3 major / 2 minor

Summary. The paper develops a basic framework for performing optimization via gradient descent on weak Riemannian manifolds (modeled on non-Banach spaces), introduces the notion of a Hesse manifold, and establishes foundational properties (including well-defined descent directions and gradient flows) for several classes of such manifolds arising in shape analysis and shape optimization.

Significance. If the central claims hold, the work would supply a missing theoretical bridge between weak Riemannian geometry and practical optimization algorithms on infinite-dimensional shape spaces, where strong Riemannian or Hilbert structures are often unavailable. The explicit construction of Hesse manifolds and the verification on concrete shape-analysis examples would be a substantive contribution to the literature on non-Banach optimization.

major comments (3)

[§3.1] §3.1, Definition 3.4: the weak metric g is asserted to induce a continuous isomorphism T*M → TM so that grad f lies in the tangent space, yet the proof only verifies injectivity and does not establish surjectivity or continuity of the inverse on the dual; without this, the gradient flow equation (3.7) is not intrinsically defined on the manifold.
[§5.2] §5.2, Theorem 5.3: the claim that the L2-type weak metric on the space of immersions yields a Hesse manifold relies on an a-priori regularity assumption that the metric is coercive on the tangent space; this reduces the setting to a strong Riemannian manifold and contradicts the paper’s emphasis on genuinely weak structures.
[§4.1] §4.1, Proposition 4.2: the existence of a well-defined descent direction for the energy functional is shown only formally; no estimate is given controlling the difference between the weak gradient and its distributional counterpart, leaving open whether the flow remains on the manifold for positive time.

minor comments (2)

[§2.3] Notation for the dual pairing in §2.3 is inconsistent with the tangent-bundle identification used later; a single global symbol would improve readability.
[Figure 1] Figure 1 caption does not indicate the precise weak metric used for the plotted trajectories; adding this detail would help readers reproduce the numerical example.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough review and valuable suggestions. We address each major comment below and have made revisions to clarify and strengthen the arguments where necessary.

read point-by-point responses

Referee: [§3.1] §3.1, Definition 3.4: the weak metric g is asserted to induce a continuous isomorphism T*M → TM so that grad f lies in the tangent space, yet the proof only verifies injectivity and does not establish surjectivity or continuity of the inverse on the dual; without this, the gradient flow equation (3.7) is not intrinsically defined on the manifold.

Authors: We agree that the original proof was incomplete. We have revised the manuscript by adding a detailed proof of surjectivity and continuity of the inverse map in the updated Definition 3.4 and Lemma 3.5. This ensures that the gradient is well-defined in the tangent space and the flow equation is intrinsically defined. revision: yes
Referee: [§5.2] §5.2, Theorem 5.3: the claim that the L2-type weak metric on the space of immersions yields a Hesse manifold relies on an a-priori regularity assumption that the metric is coercive on the tangent space; this reduces the setting to a strong Riemannian manifold and contradicts the paper’s emphasis on genuinely weak structures.

Authors: We appreciate this observation. The coercivity assumption is used for the specific example to guarantee the existence of the gradient, but the manifold itself remains weak Riemannian as the metric is not equivalent to a Hilbert structure on the full tangent space. We have added a clarifying remark in §5.2 explaining that this does not reduce the general framework to strong Riemannian manifolds, while acknowledging the limitation for this example. revision: partial
Referee: [§4.1] §4.1, Proposition 4.2: the existence of a well-defined descent direction for the energy functional is shown only formally; no estimate is given controlling the difference between the weak gradient and its distributional counterpart, leaving open whether the flow remains on the manifold for positive time.

Authors: The referee is correct that additional estimates are required. In the revised version, we have included a new estimate in Proposition 4.3 that bounds the difference between the weak gradient and the distributional gradient, ensuring that the flow stays within the manifold for positive time under suitable regularity conditions on the initial data. revision: yes

Circularity Check

0 steps flagged

No circularity: framework rests on standard differential geometry without self-referential reductions

full rationale

The paper introduces a framework for gradient descent on weak Riemannian manifolds (modeled on non-Banach spaces) and defines Hesse manifolds, with properties for shape-analysis classes. No quoted equations or definitions reduce a claimed prediction or result to a fitted input or prior self-citation by construction. The abstract and described claims rely on external Riemannian geometry background rather than internal redefinitions or ansatzes smuggled via self-citation. The central notion of well-defined gradient flow is presented as following from the weak metric structure without the derivation chain looping back to its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract alone supplies insufficient detail to enumerate free parameters, axioms, or invented entities; typical background from differential geometry (e.g., manifold structure, tangent spaces) is assumed but not listed.

pith-pipeline@v0.9.0 · 5350 in / 1031 out tokens · 26983 ms · 2026-05-15T00:33:21.696815+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Definition 2.4 (Riemannian Gradient). ... D f(p)(v) = g_p(∇f(p),v) ... Definition 3.5. A weak Riemannian C^∞-manifold (M,g) is called a Hesse manifold if it admits a metric spray S_g.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Theorem 1.4. Every robust Riemannian C^∞-manifold (M,g) is a Hesse manifold. ... Proposition 6.6. ... L²-metric ... is a robust Riemannian metric

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · 1 internal anchor

[1]

Absil, R

P.-A. Absil, R. Mahony, and R. Sepulchre. Optimization algorithms on matrix manifolds.Princeton University Press, 2008

work page 2008
[2]

Altschuler, S

J. Altschuler, S. Chewi, P. R. Gerber, and A. Stromme. Averaging on the Bures-Wasserstein manifold: dimension-free convergence of gradient descent. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan, editors,Advances in Neural Information Processing Systems, volume 34, pages 22132–22145. Curran Associates, Inc., 2021

work page 2021
[3]

Amiri, H

H. Amiri, H. Glöckner, and A. Schmeding. Lie groupoids of mappings taking values in a Lie groupoid. Arch. Math. (Brno), 56(5):307–356, 2020

work page 2020
[4]

Balehowsky, C.-J

T. Balehowsky, C.-J. Karlsson, and K. Modin. Shape analysis via gradient flows on diffeomorphism groups.Nonlinearity, 36(2):862–877, 2023

work page 2023
[5]

Bauer, M

M. Bauer, M. Bruveris, and P. W. Michor. Overview of the geometries of shape spaces and diffeomor- phism groups.J. Math. Imaging Vis., 50(1-2):60–97, 2014

work page 2014
[6]

Bauer, N

M. Bauer, N. Charon, E. Klassen, S. Kurtek, T. Needham, and T. Pierron. Elastic metrics on spaces of Euclidean curves: theory and algorithms.J. Nonlinear Sci., 34(3):38, 2024. Id/No 56

work page 2024
[7]

Bauer, C

M. Bauer, C. Maor, and B. Wirth. Completeness of reparametrization-invariant sobolev metrics on the space of surfaces, 2025

work page 2025
[8]

Borchard and G

N. Borchard and G. Wachsmuth. Characterization of Hilbertizable spaces via convex functions. Preprint, arXiv:2506.04686 [math.FA] (2025), 2025

work page arXiv 2025
[9]

Boumal.An introduction to optimization on smooth manifolds

N. Boumal.An introduction to optimization on smooth manifolds. Cambridge University Press, 2023

work page 2023
[10]

S. Chen, S. Ma, A. Man-Cho So, and T. Zhang. Proximal gradient method for nonsmooth optimization over the Stiefel manifold.SIAM Journal on Optimization, 30(1):210–239, 2020

work page 2020
[11]

E.DöhrerandN.Freches.Convergenceofgradientflowsonknottedcurves.Preprint, arXiv:2511.07214 [math.CA] (2025), 2025

work page arXiv 2025
[12]

H. I. Elíasson. Condition (C) and geodesics on Sobolev manifolds.Bull. Am. Math. Soc., 77:1002–1005, 1971

work page 1971
[13]

H. I. Eliasson. Convergence of gradient curves on Hilbert manifolds.Math. Z., 136:107–116, 1974

work page 1974
[14]

P. M. N. Feehan. On the Morse-Bott property of analytic functions on Banach spaces with Łojasiewicz exponent one half.Calc. Var. Partial Differ. Equ., 59(2):50, 2020. Id/No 87

work page 2020
[15]

P. M. N. Feehan and M. Maridakis. Łojasiewicz-simon gradient inequalities for analytic and Morse- Bott functions on Banach spaces.J. Reine Angew. Math., 765:35–67, 2020

work page 2020
[16]

Gage and R

M. Gage and R. S. Hamilton. The heat equation shrinking convex plane curves.J. Differ. Geom., 23:69–96, 1986

work page 1986
[17]

What is something (non-trivial) that can be done in Hilbert space but not Banach spaces for optimization problems? Mathematics Stack Exchange

gerw (https://math.stackexchange.com/users/58577/gerw). What is something (non-trivial) that can be done in Hilbert space but not Banach spaces for optimization problems? Mathematics Stack Exchange. URL:https://math.stackexchange.com/q/3279480 (version: 2019-07-01)

work page arXiv 2019
[18]

Greub, S

W. Greub, S. Halperin, and R. Vanstone.Connections, curvature, and cohomology. Vol. II: Lie groups, principal bundles, and characteristic classes, volume 47 ofPure Appl. Math., Academic Press. Aca- demic Press, New York, NY, 1973

work page 1973
[19]

D. W. Henderson. Infinite-dimensional manifolds are open subsets of Hilbert space.Topology, 9:25–33, 1970

work page 1970
[20]

Jost.Riemannian geometry and geometric analysis

J. Jost.Riemannian geometry and geometric analysis. Universitext. Cham: Springer, 7th edition edition, 2017

work page 2017
[21]

W. P. A. Klingenberg.Riemannian geometry, volume 1 ofDe Gruyter Stud. Math.Berlin: Walter de Gruyter, 2nd ed. edition, 1995

work page 1995
[22]

Kressner, M

D. Kressner, M. Steinlechner, and B. Vandereycken. Low-rank tensor completion by Riemannian optimization.BIT, 54(2):447–468, June 2014

work page 2014
[23]

Kristel and A

P. Kristel and A. Schmeding. The Stacey-Roberts lemma for Banach manifolds.SIGMA, Symmetry Integrability Geom. Methods Appl., 21:paper 037, 20, 2025

work page 2025
[24]

Lang.Fundamentals of differential geometry., volume 191 ofGrad

S. Lang.Fundamentals of differential geometry., volume 191 ofGrad. Texts Math.New York, NY: Springer, corr. 2nd printing edition, 2001

work page 2001
[25]

J. M. Lee.Riemannian manifolds: an introduction to curvature, volume 176 ofGrad. Texts Math. New York, NY: Springer, 1997

work page 1997
[26]

Loayza-Romero, L

E. Loayza-Romero, L. Pryymak, and K. Welker. A Riemannian approach for PDE constrained shape optimization over the diffeomorphism group using outer metrics. Preprint, arXiv:2503.22872 [math.OC] (2025), 2025

work page arXiv 2025
[27]

Loayza-Romero and K

E. Loayza-Romero and K. Welker. Numerical techniques for geodesic approximation in Riemannian shape optimization. Preprint, arXiv:2504.01564 [math.OC] (2025), 2025

work page arXiv 2025
[28]

J. Lott. Some geometric calculations on Wasserstein space.Commun. Math. Phys., 277(2):423–437, 2008. 28 V ALENTINA ZALBERTUS, MAX PFEFFER, AND ALEXANDER SCHMEDING

work page 2008
[29]

Micheli, P

M. Micheli, P. W. Michor, and D. Mumford. Sobolev metrics on diffeomorphism groups and the derived geometry of spaces of submanifolds.Izv. Ross. Akad. Nauk Ser. Mat., 77(3):109–138, 2013

work page 2013
[30]

P. W. Michor.Manifolds of differentiable mappings, volume 3 ofShiva Math. Ser.Shiva Publishing Limited, Nantwich, Cheshire, 1980

work page 1980
[31]

P. W. Michor. Manifolds of mappings and shapes. Preprint, arXiv:1505.02359 [math.DG] (2015), 2015

work page internal anchor Pith review Pith/arXiv arXiv 2015
[32]

P. W. Michor and D. Mumford. An overview of the Riemannian metrics on spaces of curves using the Hamiltonian approach.Appl. Comput. Harmon. Anal., 23(1):74–113, 2007

work page 2007
[33]

F. Otto. The geometry of dissipative evolution equations: The porous medium equation.Commun. Partial Differ. Equations, 26(1-2):101–174, 2001

work page 2001
[34]

R. S. Palais. Morse theory on Hilbert manifolds.Topology, 2:299–340, 1963

work page 1963
[35]

R. S. Palais.Foundations of global non-linear analysis. Math. Lect. Note Ser. The Ben- jamin/Cummings Publishing Company, Reading, MA, 1968

work page 1968
[36]

R. S. Palais and S. Smale. A generalized Morse theory.Bull. Am. Math. Soc., 70:165–172, 1964

work page 1964
[37]

Rudin.Real and complex analysis.New York, NY: McGraw-Hill, 3rd ed

W. Rudin.Real and complex analysis.New York, NY: McGraw-Hill, 3rd ed. edition, 1987

work page 1987
[38]

Schmeding.An introduction to infinite-dimensional differential geometry, volume 202 ofCamb

A. Schmeding.An introduction to infinite-dimensional differential geometry, volume 202 ofCamb. Stud. Adv. Math.Cambridge: Cambridge University Press, 2023

work page 2023
[39]

Schrader, G

P. Schrader, G. Wheeler, and V.-M. Wheeler. On theH1pdsγq-gradient flow for the length functional. J. Geom. Anal., 33(9):49, 2023. Id/No 297

work page 2023
[40]

Si, P.-A

W. Si, P.-A. Absil, W. Huang, R. Jiang, and S. Vary. A Riemannian proximal Newton method.SIAM Journal on Optimization, 34(1):654–681, 2024

work page 2024
[41]

Smyrlis and V

G. Smyrlis and V. Zisis. Local convergence of the steepest descent method in Hilbert spaces.J. Math. Anal. Appl., 300(2):436–453, 2004

work page 2004
[42]

A. Trouvé. Diffeomorphisms groups and pattern matching in image analysis.Commun. Partial Differ. Equations, 28(3):213–221, 1998

work page 1998
[43]

T. T. Truong. Some iterative algorithms on Riemannian manifolds and Banach spaces with good global convergence guarantee. Preprint, arXiv:2505.22180 [math.OC] (2025), 2025

work page arXiv 2025
[44]

Younes.Shapes and diffeomorphisms, volume 171 ofAppl

L. Younes.Shapes and diffeomorphisms, volume 171 ofAppl. Math. Sci.Berlin: Springer, 2nd updated edition edition, 2019. Georg-August-University Göttingen, Institute for Applied and Numerical Mathematics, Lotzestr. 16-18, 37083 Göttingen Email address:v.zalbertus@stud.uni-goettingen.de Georg-August-University Göttingen, Institute for Applied and Numerical ...

work page 2019