REVIEW 3 major objections 5 minor 7 references
Approximation of starshaped sets using polynomials
T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Polystar bodies are uniformly dense in continuous starbodies, and their degree-d approximations attain the optimal worst-case error rate.
desk verdict A solid new framework for approximating starshaped sets with polynomials; two fixable typos in the proof details shouldn't sink a paper that otherwise contributes a real tool. 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 spherical convolution operator $T_u(f)(x)=\int_{S^{n-1}} u(\langle x,y\rangle) f(y)\,d\mu(y)$ with a nonnegative univariate polynomial filter $u$. For the approximation theorem the filter is $u_{2k}(t)=c\bigl(C_{k+1}^{(\alpha)}(t)/(t-\lambda_{k+1})\bigr)^2$, whose defining property is $T_{u_{2k}}(\langle x,\cdot\rangle)(x)=\lambda_{k+1}$, the largest root of the Gegenbauer polynomial of degree $k+1$; this identity converts the bound on squared geodesic distance into the explicit error constant. The same operator preserves support functions, which makes the polygauge approximation of a convex body convex. On the algorithmic side, the Funk–Hecke formula decomposes $T_u$ into a scaling of each spherical-harmonic component, and an exact quadrature rule approximates each component from finitely many evaluations. The opposite direction is carried by a Kolmogorov-width argument: a packing of $2^{N+1}$ tent-like Lipschitz functions on the sphere forces any $N$-dimensional subspace to have worst-case error at least $\kappa\beta/2$.
What would settle it
Run Algorithm 2 on the cube's radial function $\rho(x)=\min_i 1/|x_i|$ in dimension 3 (known Lipschitz constant $\kappa=1$) for degrees $d=10,20,40,80$ and compare the measured sup error against the bound $(\pi/\sqrt{2})\sqrt{1-\lambda_{d/2+1}}$; if the error is not bounded by this expression asymptotically, or if it decays slower than $C/d$ for all large $d$, then Theorem 1.1(2)'s error estimate or its constant is wrong.
Extended reading notes
Core claim
The central discovery is that the two functions defining any reasonably regular starbody—gauge and radial—can be replaced, uniformly and at the optimal asymptotic rate, by polynomials on the sphere without losing geometric information. The approximating bodies, called polystar bodies, are produced by a spherical convolution operator whose filter is a nonnegative univariate polynomial $u_{2k}(t)=c\bigl(C_{k+1}^{(\alpha)}(t)/(t-\lambda_{k+1})\bigr)^2$, where $\lambda_{k+1}$ is the largest root of the Gegenbauer polynomial of degree $k+1$ and $\alpha=(n-2)/2$. The filter is chosen so the convolution's first moment is exactly $\lambda_{k+1}$, converting a Cauchy–Schwarz estimate on geodesic distance into the explicit error bound. Optimality is stated through a Kolmogorov-width lower bound: any subspace $W$ containing the constants has worst-case error at least $\kappa\beta/2$ over the class of $\kappa$-Lipschitz positive functions, where $\beta=\min\bigl((\dim W-1)/(n-1),\,2/\kappa\bigr)$, yielding a $C_0\kappa/d$ lower bound for polynomials of degree $d$. Computationally, an exact quadrature rule together with the Funk–Hecke formula turns the convolution into a finite sum, and nonnegativity of the filter makes the output a sum of squares in the coordinate ring of the sphere.
Load-bearing premise
The $O(\kappa/d)$ guarantee requires the gauge or radial function to have a known finite Lipschitz constant $\kappa$, which the paper tries to derive from the assumption that the body is starshaped with respect to a whole ball; for the radial function that derivation as printed divides by the sup of $\gamma^2$ instead of the inf, so the stated $R^2/r$ constant is not established by the displayed argument.
Editorial extensions
If this is right
- Density: every starbody with continuous gauge or radial function can be uniformly approximated by polystar bodies, so the class is rich enough to model all continuous star-shaped geometry.
- Rate: for Lipschitz starbodies a degree-$d=2k$ polystar approximation exists with sup error at most $(\pi/\sqrt{2})\sqrt{1-\lambda_{k+1}}\,\kappa \sim \pi(n-2)\kappa/(\sqrt{2}\,d)$, and the paper gives geometric sufficient conditions: starshapedness around a ball of radius $r$ inside $B_R(0)$ yields $\kappa=1/r$ for the gauge and $\kappa=R^2/r$ for the radial function.
- Optimality: no finite-dimensional subspace of dimension equal to that of degree-$d$ polynomials achieves an asymptotically smaller worst-case error on the class of $\kappa$-Lipschitz functions, so the polynomial model is dimension-optimal.
- Computability: the approximating polynomial can be constructed from black-box evaluations of the gauge/radial function using quadrature rules with polynomially many nodes, and the result is a sum of squares on the sphere, making subsequent maximization semidefinite-representable.
- Geometry preservation: the intersection body of a polyradial body is polyradial, and polygauge approximations of convex bodies are convex, so the method transfers algebraic structure to geometric invariants like largest-volume slices and widths.
Reading between the lines
- Algorithm 1 is a general recipe for uniform polynomial approximation of any black-box Lipschitz function on the sphere, independent of starbodies; it could be reused for spherical designs, quadrature construction, or other sphere-constrained optimization tasks.
- The node count of the proposed quadrature rules grows like $(k+1)^{n-1}$, so the polynomial-time statement is for fixed dimension; extending the method to the slicing-conjecture regime $n\ge 5$ will require sparse or adaptive quadrature.
- The paper separates approximation error from optimization error: after a polystar approximation is built, the remaining cost is maximizing a polynomial on the sphere, where sum-of-squares hierarchies converge at $O(1/\ell)$ or $O(1/\ell^2)$; for high degrees this optimization step, not approximation, is likely the practical bottleneck.
- Because the convolution filter is nonnegative, the output is a sum of squares, so the same geometric approximation pipeline could be adapted to certify positivity of gauge/radial-like functions on spheres, not just to represent them.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces polystar bodies, defined as starbodies whose gauge or radial function agrees with a polynomial on the unit sphere, and develops a theory of polynomial approximation for starshaped geometry. The main results are: density of polygauge and polyradial bodies among continuous starbodies; quantitative uniform approximation of Lipschitz gauge/radial functions by degree-d polystar bodies with error O(kappa/d), based on spherical convolution with nonnegative Gegenbauer filters; a Kolmogorov-width lower bound showing this rate is asymptotically optimal; constructive quadrature-based algorithms for computing such approximations from black-box evaluations; and numerical demonstrations for intersection bodies, largest-volume slices, and widths of several three-dimensional bodies.
Significance. If the results hold, polystar bodies provide a computational model class for starshaped geometry that is both dense and optimally accurate in a worst-case sense. The paper's strengths are the constructive nature of the approximation theorem with explicit constants obtained from Gegenbauer root bounds, the absence of fitted parameters in the error estimates, and the numerical validation against independently known exact values such as the cube's largest slice direction and the octahedron's width. The algorithms require only black-box evaluations and quadrature, and the sum-of-squares structure in Theorem 4.6 is attractive for subsequent optimization. These contributions are substantial, and the main ideas are sound and publishable after the errors identified below are corrected.
major comments (3)
- [3.3, proof of Theorem 1.1(3)] The inequality displayed in the proof, |rho(y)-rho(x)| <= |gamma(y)-gamma(x)| / sup_z |gamma(z)|^2, is written in the wrong direction: since gamma(x)gamma(y) can be much smaller than sup gamma^2, dividing by the supremum gives a lower bound rather than an upper bound. The proof as printed therefore yields the constant m/r with m = inf rho^2, which is not generally a valid Lipschitz constant for rho, and it does not establish the theorem's stated constant R^2/r. The claim R^2/r is nevertheless true and can be proved by the corrected estimate |rho(y)-rho(x)| = |gamma(y)-gamma(x)|/(gamma(x)gamma(y)) <= (1/(inf gamma)^2)|gamma(y)-gamma(x)| <= R^2 * (1/r) d(x,y), using rho <= R. This step is load-bearing for the abundance-of-Lipschitz-starbodies assertion and must be repaired.
- [3.4, Theorem 1.2] The statement defines beta = min(dim(W) - 1/(n-1), 2/kappa), which is not dimensionally meaningful and does not match the proof. The proof requires beta to satisfy N beta^{n-1} < 1 with N = dim W, and also beta < 2/kappa so that delta = beta/2 satisfies the positivity condition in Lemma 3.5. The intended definition is evidently beta = min(c_N N^{-1/(n-1)}, c_kappa/kappa) up to absolute constants from the volume argument. The later dimensional passage of the proof, including the corrupted expression 'beta < N - 1/n+1', also needs a clean restatement. As printed, the lower bound kappa beta/2 depends on an ill-defined quantity, so the theorem cannot be checked in its current form.
- [4.2, Algorithm 2; 5.1, Algorithm 3] The zonal expansions in Algorithm 2 are summed only over j = 1, ..., d, and in Algorithm 3 only over even j starting at j = 2. In both cases the j = 0 harmonic component is omitted. For a generic positive gauge or radial function the zeroth component is nonzero, and omitting it introduces an O(1) error that does not vanish with d; this invalidates the stated uniform approximation guarantee in Algorithm 2 and the intersection-body approximation formula in Algorithm 3. The fix is to include the j = 0 term (with the corresponding Radon-transform multiplier in Algorithm 3) or to compute and add the constant component separately.
minor comments (5)
- [2.2, Proposition 2.7(2)] The sentence 'The body L1 (resp. L2) is convex if and only if its polar body is convex' is false as stated, because the polar of any set is convex by definition. The intended equivalence is between convexity of L and the function p (resp. 1/q) being a support function; the polar-body sentence should be removed or reformulated.
- [2.2, proof of Proposition 2.7(1)] The text says 'using the fact that L1 is polyradial' in the argument about L1; this should say 'polygauge'.
- [4.1, Algorithm 1] Algorithm 1 says the quadrature rule is 'exact in degrees d+m and 2d'; this should read 'exact in degree max(d+m, 2d)', matching Proposition 4.3.
- [3.4, Lemma 3.5 and Theorem 1.2 proof] The proof of Theorem 1.2 switches notation from f_tau to rho_{L*} midway; since f_tau is used as a gauge function, the notation should be made consistent.
- [2.1] The claim that every positive function on S^{n-1} is the gauge/radial function of some starbody requires the function to be continuous (or at least measurable with suitable boundedness) for the resulting set to be compact with interior containing the origin; this regularity condition should be stated.
Circularity Check
No significant circularity: the approximation bounds are derived from external theorems (Newman–Shapiro, Ragozin, Driver–Jordaan) and carry no fitted parameters or self-citation-dependent premises.
full rationale
The paper's central claims are self-contained mathematical deductions. Theorem 3.1 is proved in the text from the explicit filter u_2k, the Newman–Shapiro bound arccos^2(t) <= (pi^2/2)(1-t), Gegenbauer orthogonality, and the Driver–Jordaan root bound; no parameter is fitted to the target starbodies, and the constants depend only on the Lipschitz constant kappa, dimension, and degree. Theorem 1.1 then applies that theorem to gauge/radial functions and constructs the approximating polystar body directly from the polynomial T_u(f_L), so the error estimate is not an input renamed as a prediction. Theorem 1.2 is an independent Kolmogorov-width lower bound built from explicit Lipschitz functions and a volume argument; it does not assume the result it proves. Numerical examples are checked against independently known values (cube slice direction from Ball, octahedron width, cylinder/tin-can geometry), so the computations are not validated by the paper's own outputs. The few author self-citations ([CV24], [BBMS22], [BLM23]) are used for an explicit quadrature construction, for known intersection-body descriptions of polytopes, and as motivational context; none of these supplies the load-bearing approximation theorem, and each cited result is independently checkable. The proof of Theorem 1.1(3) contains a local inequality-direction issue in the radial Lipschitz constant estimate, but that is a correctness defect in a constant bound, not a circular reduction: it does not make the conclusion equivalent to an input.
Assumptions & free parameters
assumptions (5)
- standard math Lipschitz continuity of the gauge function is equivalent to starshapedness with respect to a ball of radius r around the origin, with 1/r a valid Lipschitz constant.
- standard math The spherical convolution approximation framework of Newman-Shapiro and Ragozin gives O(1/d) uniform approximation for Lipschitz functions on the sphere.
- standard math Kiderlen's theorem: convolution with a nonnegative kernel preserves support functions, hence preserves convexity of the associated body.
- standard math Funk-Hecke formula, Gegenbauer orthogonality, and Gaussian quadrature exactness on [-1,1].
- standard math Stone-Weierstrass density of polynomials restricted to the sphere in the space of continuous functions.
invented entities (1)
-
Polystar bodies
Cite this review
Pith. "Pith review of Approximation of starshaped sets using polynomials." pith.science (2026). https://pith.science/paper/UOZBAU7M
@misc{pith2026250524352,
author = {Pith},
title = {Pith review of: Approximation of starshaped sets using polynomials},
year = {2026},
howpublished = {\url{https://pith.science/paper/UOZBAU7M}},
note = {Machine review of arXiv:2505.24352}
}
read the original abstract
We introduce polystar bodies: compact starshaped sets whose gauge or radial functions are expressible by polynomials, enabling tractable computations, such as that of intersection bodies. We prove that polystar bodies are uniformly dense in starshaped sets and obtain asymptotically optimal approximation guarantees. We develop tools for the construction of polystar approximations and illustrate them via several computational examples, including numerical estimations of largest volume slices and widths.
Figures
Figures from the paper (5 more)
Reference graph
Works this paper leans on
-
[5]
[KP03] Stefan Kunis and Daniel Potts,Fast spherical Fourier algorithms, J. Comput. Appl. Math.161(2003), no. 1, 75–98. [Lor60] George G. Lorentz,Lower bounds for the degree of approximation, Transactions of the American Mathe- matical Society97(1960), 25–34. [NS64] Donald J. Newman and Harold S. Shapiro,Jackson ’s theorem in higher dimensions, On Approxim...
work page 2003
-
[6]
Ragozin,Polynomial approximation on compact manifolds and homogeneous spaces, Trans
[Rag70] David L. Ragozin,Polynomial approximation on compact manifolds and homogeneous spaces, Trans. Amer. Math. Soc.150(1970), 41–53. [Rag71] ,Constructive polynomial approximation on spheres and projective spaces, Trans. Amer. Math. Soc.162(1971), 157–170. [Rag72] ,Uniform convergence of spherical harmonic expansions, Math. Ann.195(1972), 87–94. [Rez95...
work page 1970
-
[1978]
Reprinting of the 1962 original. MR542284 [WR24] Inc. Wolfram Research,Mathematica,
work page 1962
-
[2001]
MR1805196 [Bal86] Keith Ball,Cube slicing inR n, Proc. Amer. Math. Soc.97(1986), no. 3, 465–473. [BBMS22] Katalin Berlow, Marie-Charlotte Brandenburg, Chiara Meroni, and Isabelle Shankar,Intersection bodies of polytopes, Beitr. Algebra Geom.63(2022), no. 2, 419–439. [BCS23] Abdesslam Boulkhemair, Abdelkrim Chakib, and Azeddine Sadik,On a shape derivative ...
work page 1986
-
[2013]
[Fay04] Leonid Faybusovich,Global optimization of homogeneous polynomials on the simplex and on the sphere, Frontiers in global optimization, 2004, pp. 109–121. [FF21] Kun Fang and Hamza Fawzi,The sum-of-squares hierarchy on the sphere and applications in quantum information theory, Mathematical Programming190(2021), no. 1-2, 331–360. [Fun13] Paul Funk,¨ ...
work page 2021
-
[2023]
[CV24] Sergio Cristancho and Mauricio Velasco,Harmonic hierarchies for polynomial optimization, SIAM J. Optim.34(2024), no. 1, 590–615. [DH94] James R. Driscoll and Dennis M. Healy Jr.,Computing Fourier transforms and convolutions on the 2-sphere, Adv. in Appl. Math.15(1994), no. 2, 202–250. [DJ12] Kathy Driver and Kerstin Jordaan,Bounds for extreme zeros...
work page 2024
-
[2024]
[GW69] Gene H. Golub and John H. Welsch,Calculation of Gauss quadrature rules, Math. Comp.23(1969), 221–230; addendum, ibid. 23 (1969), no. 106, loose microfiche suppl. A1–A10. MR245201 [HHMM20] Guillermo Hansen, Irmina Herburt, Horst Martini, and Maria Moszy´ nska,Starshaped sets, Aequationes Math.94(2020), no. 6, 1001–1092. [Kid06] Markus Kiderlen,Blasc...
work page 1969
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.