Energy minimisation using overlapping tensor-product free-knot B-splines

Alexandre Magueresse; Santiago Badia

REVIEW 3 major objections 2 minor 1 cited by

Energy minimisation using overlapping tensor-product free-knot B-splines

T0 review · 3 major / 2 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Free-knot B-splines satisfy the convergence conditions for joint optimisation of knots and coefficients.

desk verdict A credible but abstract-only claim that free-knot B-spline energy minimization fits into the authors' own convergence framework; the unspecified 'mild mesh size condition' is the load-bearing detail that full-text review must pin down. read the letter →

arxiv 2508.17705 v2 pith:TBZWRZ4Y submitted 2025-08-25 math.NA cs.NA

Alexandre Magueresse , Santiago Badia This is my paper · ORCID

classification math.NAcs.NA MSC 65D0765N3065K10

keywords free-knotB-splinesoverlappingpatchesenergyminimisationadaptivediscretisationellipticPDEsprojectedgradientdescentnonlinearapproximationknotpositions

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that the discrete energy minimisation problem for overlapping tensor-product free-knot B-spline patches, where knot positions are free variables, fits the abstract optimisation framework from the authors' companion work. For linear, self-adjoint elliptic PDEs, it argues that under a mild mesh size condition the discrete energy satisfies the structural properties that guarantee local and global convergence of projected gradient descent over both knot positions and coefficients. If correct, this turns a heuristic adaptive space into one with provable convergence, allowing localised PDE features to be captured with fewer degrees of freedom than standard finite elements.

What carries the argument

The central object is a finite-dimensional trial space assembled from overlapping tensor-product B-spline patches; the knot positions are not fixed a priori but act as nonlinear optimisation variables that move the discretisation toward localised features. The load-bearing step is verifying that the resulting discrete energy functional satisfies the abstract structural hypotheses of the companion projected-gradient-descent framework, which is what turns a heuristic knot-moving scheme into a provably convergent one.

What would settle it

A reproducible numerical test would solve a one-dimensional elliptic boundary-layer problem by projected gradient descent on knots and coefficients, monitoring the energy and the gradient norm; if convergence fails under the claimed mesh condition, or if the condition cannot be verified numerically, the guarantee is not effective.

Watch

Extended reading notes

Core claim

For linear, self-adjoint elliptic PDEs, the energy functional defined on overlapping tensor-product B-spline patches whose knot positions are free variables meets the structural requirements of the companion optimisation framework. Under a mild mesh size condition, this means the constrained optimisation problem of minimising the discrete energy jointly over knot positions and B-spline coefficients is locally and globally convergent when solved by projected gradient descent. Numerical experiments demonstrate that this adaptive space captures localised solution features with substantially fewer degrees of freedom than standard finite-element discretisations.

Load-bearing premise

The convergence conclusion depends on the companion optimisation framework being sound and on the mild mesh size condition being both satisfiable and checkable in practice.

Editorial extensions

If this is right

Projected gradient descent that updates knot positions and coefficients together is guaranteed to converge locally and globally for the energy minimisation problem of linear self-adjoint elliptic PDEs, provided the mild mesh size condition holds.
The adaptive free-knot B-spline space inherits the convergence guarantees of the companion framework without requiring a separate convergence analysis.
The method sidesteps the rigidity of h- and p-refinement: knots move continuously rather than following prescribed refinement rules, and no a posteriori error estimator is needed to drive adaptivity.
Numerically, localised solution features are resolved with significantly fewer degrees of freedom than standard finite element discretisations.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the mild mesh size condition can be checked from the current iterate and its constants are computable, the framework implies a practical adaptive algorithm: optimise knots and coefficients, check the condition, refine or coarsen, and repeat without an error estimator.
Because the analysis targets the energy functional rather than a particular equation, the same structural-property route might extend to other variational problems, such as nonlinear or non-symmetric PDEs with an energy, though the paper only claims linear self-adjoint problems.
The overlapping-patch construction may allow local knot movement without global reparameterisation, potentially mitigating the exponential cost of tensor-product spaces in higher dimensions.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague, quick take on arXiv:2508.17705. The new thing here is modest but honest: the authors show that a concrete overlapping tensor-product free-knot B-spline space satisfies the structural hypotheses of their companion optimization framework, so projected gradient descent over knots and coefficients inherits convergence guarantees. That is a useful bridge between abstract convergence theory and a practical adaptive discretization. The numerical experiments claim large degree-of-freedom savings over FEM for localized features, which is plausible for these methods but unverifiable from the abstract.

The main soft spot is the phrase 'under a mild mesh size condition.' That condition is the linchpin: it presumably prevents knot collisions, which change the basis rank and make the problem discontinuous. But the abstract never states whether the condition can be checked a priori, or whether its threshold depends on the unknown solution. If it does depend on the solution, the guarantee is not actionable for adaptivity. This is exactly the point a referee must press. I don't see internal inconsistency at the abstract level, and the dependency on the companion paper is that way, not circular: this paper verifies assumptions, it does not underpin the companion proof. Still, the companion framework is by the same authors and has no independent verification here, so the reader is essentially being asked to take two papers on faith.

Given the abstract-only basis, I'd send this to peer review rather than reject. The referee should ask for the explicit form of the mesh size condition, a demonstration that it can be verified before solving, and replication details for the experiments. If the full text is as coherent as the abstract suggests, it could be a solid contribution. If the condition turns out to depend on the unknown minimizer, the main theorem shrinks to a conditional statement that adaptivity cannot actually use.

Referee Report

3 major / 2 minor

Summary. The paper studies the energy minimisation problem for linear, self-adjoint elliptic PDEs discretised with overlapping tensor-product free-knot B-spline patches, where knot positions are treated as nonlinear optimisation parameters. The abstract claims that, under a mild mesh size condition, the discrete energy satisfies structural properties required for the local and global convergence of a constrained optimisation scheme developed in the authors' companion paper (arXiv:2508.17687). This is said to establish convergence of projected gradient descent for the joint optimisation of knot positions and coefficients, and numerical experiments are reported to illustrate the method's efficiency in capturing localised features with fewer degrees of freedom than standard finite elements.

Significance. If the claims are correct, the paper would provide a convergence guarantee for a flexible nonlinear approximation space, potentially offering an adaptive method that avoids the geometric rigidity of h- and p-refinement and does not require a posteriori error estimates. The emphasis on free-knot B-splines is of interest to the numerical analysis and scientific computing community. However, the significance is conditional on the soundness of the companion optimisation framework and on the mesh size condition being explicit and verifiable. The paper deserves attention if the full text substantiates these points; it offers a promising direction for PDE approximation, but the abstract alone does not establish that the central convergence theorem is actionable.

major comments (3)

[Abstract, paragraph 2] The 'mild mesh size condition' is the linchpin of the convergence guarantee but is never specified. The referee cannot determine whether the condition is a priori verifiable -- e.g., depending only on known data and the current iterate -- or whether its threshold involves unknown solution quantities or the as-yet-unknown minimising knot configuration. If the condition depends on the unknown solution, the guarantee would not be actionable in an adaptive setting. The full text must state the precise condition and explain how it is checked in practice, including in the numerical experiments.
[Abstract, paragraph 2] The convergence result is inherited directly from the authors' companion paper (arXiv:2508.17687) by verifying structural properties, but the abstract offers no independent evidence that the companion framework is sound. If the companion proof itself relies on results from the present paper, the dependency would be circular. The authors should clearly delineate the dependency structure and, at minimum, outline the structural properties verified here and how they map to the companion assumptions, providing either a proof sketch or a pointer to a machine-checked verification.
[Abstract, paragraph 3] The numerical experiments are described only qualitatively: the abstract states that they 'illustrate the method's efficiency' and capture localised features with 'significantly fewer degrees of freedom', but no quantitative results, problem setup, error metrics, or convergence rates are given. Since the empirical claims are part of the paper's support, the full text must include experiment details, including how the mesh size condition is enforced or tested, and comparisons with standard finite element methods using error versus degrees of freedom plots or tables.

minor comments (2)

[Abstract, paragraph 1] The abstract does not specify the class of linear, self-adjoint elliptic PDEs (e.g., coefficient regularity, boundary conditions, spatial dimension). Stating the scope would help readers judge the generality of the result.
[Title and abstract] The title uses British spelling ('minimisation') while the abstract mainly uses the same, but there is a mixture with 'minimising' in the final paragraph; standardise the spelling throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity demonstrable from abstract; conditional theorem and companion citation do not reduce the result to its inputs.

full rationale

Full text is not available, so this audit is limited to the abstract's derivation chain. The abstract claims that, under a mild mesh size condition, the discrete energy satisfies the structural properties required by the companion optimisation framework. That is a conditional verification of the framework's hypotheses, not an equivalence between input and output. The same-author companion citation is load-bearing for the convergence statement, but a citation to a prior theorem is not circular unless the cited theorem is itself derived from the present paper or its hypotheses include the claim being proved. No such dependency is visible in the abstract. The unspecified 'mild mesh size condition' is a verification gap: if the condition were unverifiable a priori or depended on the unknown solution, the guarantee would be weak, but vagueness of an assumption is not circularity. No equation or construction in the abstract reduces a result to its own inputs, and no fitted parameter is renamed as a prediction. Therefore no circular step can be exhibited, and the score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

Abstract-only review: no explicit fitted constants are visible in the abstract. The knot positions are the method's optimization variables, not ad hoc fitted parameters, so they are not listed as free parameters. Any hidden constant in the 'mild mesh size condition' cannot be identified from the abstract and would belong here if the full text reveals one. No invented entities are introduced; the contribution is analytical. The axioms record the load-bearing dependencies: the companion framework's correctness and the unspecified mesh size condition, both of which the abstract takes as input.

assumptions (3)

domain assumption Correctness of the companion framework (arXiv:2508.17687) and its structural-property conditions for convergence of projected gradient descent.
The abstract states the discrete energy 'satisfies the structural properties required' by that framework; the convergence guarantee is therefore inherited from a same-author prior paper whose verification status is not described. If the companion framework is flawed, the central claim collapses.
standard math Standard well-posedness and approximation theory for linear, self-adjoint elliptic PDEs in overlapping tensor-product B-spline spaces.
The energy minimization setup presupposes coercivity, continuity, and approximation properties of the B-spline space; the abstract takes these for granted without restating them.
domain assumption The 'mild mesh size condition' holds for the practical cases tested and is verifiable within the adaptive algorithm.
The abstract asserts the structural properties 'under a mild mesh size condition' but does not specify the threshold, how it depends on problem data, or whether the numerical experiments check it. If the condition is not satisfiable or not checkable, the guarantee is not actionable.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Energy minimisation using overlapping tensor-product free-knot B-splines." pith.science (2026). https://pith.science/paper/TBZWRZ4Y

@misc{pith2026250817705,
  author       = {Pith},
  title        = {Pith review of: Energy minimisation using overlapping tensor-product free-knot B-splines},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TBZWRZ4Y}},
  note         = {Machine review of arXiv:2508.17705}
}

abstract

Accurately solving PDEs with localised features requires refined meshes that adapt to the solution. Traditional numerical methods, such as finite elements, are linear in nature and often ineffective for such problems, as the mesh is not tailored to the solution. Adaptive strategies, such as $h$- and $p$-refinement, improve efficiency by sequentially refining the mesh based on a posteriori error estimates. However, these methods are geometrically rigid -- limited to specific refinement rules -- and require solving the problem on a sequence of adaptive meshes, which can be computationally expensive. Moreover, the design of effective a posteriori error estimates is problem-dependent and non-trivial. In this work, we study a specific nonlinear approximation scheme based on overlapping tensor-product free-knot B-spline patches, where knot positions act as nonlinear parameters controlling the geometry of the discretisation. We analyse the corresponding energy minimisation problem for linear, self-adjoint elliptic PDEs, showing that, under a mild mesh size condition, the discrete energy satisfies the structural properties required for the local and global convergence of the constrained optimisation scheme developed in our companion work [Magueresse, Badia (2025, arXiv:2508.17687)]. This establishes a direct connection between the two analyses: the adaptive free-knot B-spline space considered here fits into the abstract framework, ensuring convergence of projected gradient descent for the joint optimisation of knot positions and coefficients. Numerical experiments illustrate the method's efficiency and its ability to capture localised features with significantly fewer degrees of freedom than standard finite element discretisations.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Parametric Neural r-Adaptivity for Isogeometric Analysis via Residual Minimization
math.NA 2026-07 conditional novelty 6.0 of 10

A residual-based a posteriori estimator is used as a differentiable loss to train a neural network that relocates IGA knots, with a density output that works at any refinement level.

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.