CS Paper Radar

Decoy and disclosure radii of invariant shape descriptors

Tanush Shaska, Lubjana Beshaj

cs.CV astro-ph.EPcs.CRmath.AG ·submitted 2026-10-06 14:16 UTC · v1
unusual

Abstract

A recognizer that compares rotation-invariant descriptors sees a surface only up to the fiber of the descriptor. We measure this fiber by its radius in the orbit distance from the enrolled surface. A large radius admits decoys, that is, distant shapes that pass the matcher. A small radius discloses the enrolled shape to anyone who captures the stored value. For star-shaped surfaces truncated to spherical harmonics of degree at most $L$, with $n$ coefficients, a descriptor of generic rank $r$ has generic fibers of dimension $n-3-r$ modulo rotations. The standard pool of band powers, even bispectra, and three invariants of the degree-three band therefore admits decoy families of dimension $5$, $13$, $20$ at $L=4,6,8$. Its rank first reaches $n-3$ at $L=16$, and a mirror decoy remains at every $L$. The odd bispectra remove the mirror decoy generically for $L \geq 4$. Yet at fixed mean radius the same pool determines the enclosed volume exactly, and it does not determine whether a surface meets a clearance requirement. We certify two cases by exact and interval arithmetic. At $L=6$ a decoy matches all $32$ invariants to relative precision $2 \cdot 10^{-18}$ at orbit distance at least $0.87$ times the norm of the enrolled tuple. For the radar shape model of asteroid (101955) Bennu, the pool recovers the modeled volume, misses the handedness, and leaves the keep-out radius uncertain by more than $7 \, \mathrm{m}$.

Closest papers

Nearest neighbours by meaning across the whole index.

Invariant Shape Analysis of Surfaces with Spherical Topology

cs.CV math.AG ·2026-09-30 ·Tanush Shaska
theory similarity 0.54
Abstract

Every real spherical harmonic of even degree is congruent to its mirror image. We prove that a real spherical harmonic of odd degree $n\ge3$ with distinct Maxwell axes is congruent to its mirror image if and only if the hyperelliptic curve of genus $n-1$ defined by its binary form has a reduced automorphism of even order whose lift acts on the form by $-1$. We then determine the locus of these curves: it is the locus of curves with an extra involution if and only if $n-1$ is a power of two, and otherwise it also contains, for each even divisor $k$ of $n-1$ with odd cofactor $d\ge3$, an irreducible family $y^2=x K(x^{2k})$, $°K=d$, with generic reduced automorphism group $\Z/2k$. For $n=3$ the condition is the vanishing of the skew invariant of the binary sextic. As an application, the invariants of the truncated spherical harmonic expansion of a genus-zero surface, in a conformal parameterization normalized by its conformal barycenter, give a descriptor that is exactly invariant under parameterization, position and scale, records handedness, and is stable in the orbit distance.

0.49

On the Kissing Number of Craig's Lattice and Spherical List Decoding

cs.IT ·2026-11-17 ·Léo Ducas, Julian Nowakowski, Ailsa Robertson
theory similarity 0.48
Abstract

Craig's lattices $A_{p-1}^k$ are a family of lattices with large kissing numbers, i.e., relative to their dimension $n = p-1$, they contain a large number of shortest nonzero vectors. Specifically, for small $k = o(\log p / \log \log p)$, a lower bound of Bacher (2015) shows that $A_{p-1}^k$ has a super-polynomial kissing number of $ω(n^k)$. Craig's lattices are therefore attractive candidates for the construction of spherical codes. In this work, we introduce a polynomial-time list decoder for spherical codes obtained from Craig's lattices for angular distance below $45^\circ$. Our second main result shows that Bacher's lower bound is asymptotically tight for $k = o(\log p / \log \log p)$. Finally, we introduce an efficient algorithm for computing the kissing number of $A_{p-1}^k$. Using this algorithm, we compute their kissing number in dimensions up to 521. Based on this data, we pose the open question whether the kissing number of $A_{p-1}^k$ with $k \approx \sqrt{p}$ grows like $e^{Θ(\sqrt{n})}$ in dimension $n=p-1$. Concurrent work by Laarhoven and Kominers as well as Bennett, Hhan, Li and Stephens-Davidowitz proves such a bound for very closely related lattices.

0.15

An $O(L^3)$ Spherical Harmonic Transform on $L^2$ Samples

eess.SP math.NA ·2026-11-12 ·Zubair Khalid
theorycode/data similarity 0.48
Abstract · 13 pages, 7 figures, 1 table. Code: https://github.com/zubairkhalid1/osht

For the representation of signals on the sphere band-limited at $L$ using the optimal number of samples, $L^2$, equal to the number of degrees of freedom in harmonic space and placed on $L$ iso-latitude rings, we develop a spherical harmonic transform (SHT) that is theoretically exact, requires no solution of linear systems and no precomputation beyond the placement of the rings, and has complexity $O(L^3)$. The transform rests on the observation that, for each order $m$, the samples on the rings that resolve order $m$ are a complete set of samples for the order-$m$ profile of the signal, so that its values on the remaining rings follow from a closed-form Lagrange basis, the sampling functions of order $m$. We prove that the proposed optimal-dimensionality SHT (OSHT) recovers every band-limited signal exactly in exact arithmetic, so that the $L^2$ samples admit a sampling theorem in the collocation sense. For the final harmonic analysis we give two exact back-ends: Gauss-Legendre quadrature of the interpolated profiles and the algorithm of the equiangular sampling theorem, which applies without change because the rings coincide with its grid. We identify the amplification constant that governs numerical error, show that it replaces the condition number used previously for the placement of rings, propose a placement that minimizes it, and give an error propagation model that explains the growth of error with the band-limit. A multi-pass refinement converges in two passes to the rounding floor of the samples. We also describe a stable log-domain evaluation of the sampling functions, a rescaled Legendre recurrence, and a forbidden-region skipping rule that halves the cost of the elimination. Numerical experiments up to $L=2048$ characterize the accuracy, measured by the maximum and mean coefficient errors and the signal-to-noise ratio, the stability and the computation time.

0.35

The covering radius of RM(1,9) is precisely 242

cs.IT math.CO ·2026-10-15 ·Xiang-Dong Hou, Evelyn Sabo, Freeman Slaughter
code/data similarity 0.46
Abstract · 9 pages

Let $\text{RM}(r,n)$ be the $r$th order Reed-Mullet code of length $2^n$, and let $ρ(r,n)$ denote its covering radius. It is known that $242\leρ(1,9)\le 244$, and the determination of $ρ(1,9)$ is a long-standing open question. We give a computer-assisted proof that $ρ(1,9)=242$, leveraging the classification of the $349$ orbits in $\text{RM}(3,9)/\text{RM}(2,9)$ under $\text{GL}(9,2)$ [1,3]. The notebook to verify the computer search portion, Verify_RM(1,9).ipynb, contains all data and recomputes each check, and may be found in the repository https://github.com/FreemanSlaughter/Covering-Radius.

0.67

Non-Uniqueness in the Coordinate-Based Normalization of Tetrahedral Similarity Classes

cs.DM ·2026-11-09 ·Miguel Ángel Padrón, Agustín Trujillo-Pino, Jose Pablo Suárez
similarity 0.46
Abstract

Recently, a new coordinate-based normalization has been introduced in [1] to represent tetrahedral similarity classes by normalized coordinates, enabling the study of LEB orbits as a discrete dynamical system. However, the uniqueness of this representation is not guaranteed when two or more edges adjacent to the selected longest edge have the same minimum length and remain admissible under the normalization criteria. In this paper, we demonstrate that different admissible choices may lead to distinct normalized coordinate representations of the same tetrahedral similarity class. We first present an explicit counterexample in which two admissible vertex labelings yield different normalized coordinate representations related by reflection, despite corresponding to the same similarity class. We then provide a systematic classification of the nine possible edge configurations arising in the selection of the canonical face when the longest edge of the tetrahedron is unique, and analyze the corresponding geometric and algebraic conditions. Our analysis identifies the configurations in which the normalization is non-unique and distinguishes them from those in which the ambiguity is resolved by the length criterion, angle criterion, or geometric symmetries. We further show that this non-uniqueness leads to an overcounting of distinct similarity classes.

0.25

Geometric Obstructions and Sharp Genus Dependence in LCD Scalings

cs.IT math.AG ·2026-11-03 ·Rahim Rahmati-Asghar
theory similarity 0.45
Abstract · All comments are welcome

We study the least pole order of a rational function whose nonzero evaluations scale a prescribed code to a Euclidean LCD code. A residue-one differential yields an obstruction for odd-dimensional one-point AG codes in characteristic two: for $C=C_L(D,aQ)$ with $0\le a<n$ and odd dimension, the threshold is at least $n+2g-a-1$. The obstruction uses alternating Gram matrices and does not require the scaled code to be self-orthogonal. For the dimension-three codes $C_L(D,(r+1)Q)$ on Hermitian curves over $\F_{r^2}$, $r=2^e\ge4$, we obtain \[ n+2g-r-2\leτ_D(C;Q)\le n+2g-3. \] Consequently the leading genus contribution is asymptotically sharp even in dimension three. On the elliptic curve $y^2+y=x^3$ over $\F_4$, we prove the exact dimension-three threshold $6$ by an explicit multiplier. For repetition codes on the Hermitian family, the threshold equals the general interpolation bound $n+2g-1$. The upper bounds use information sets and hold for $q>3k$, and for $q>2k$ in characteristic two.

0.07

Efficient lifts into odd dimensions of Spherical $5$- and $7$-designs with Balanced Halves

cs.IT math.CO ·2026-10-08 ·Konstantin Delchev, Boryan Yordanov
similarity 0.45
Abstract · 13 pages, 3 tables

We propose constructions of spherical $5$- and $7$-designs in odd dimensions by lifting balanced halves of even-dimensional designs to several layers of suitable latitudes. For every odd dimension, we obtain an explicit spherical $5$-design with at most $60(d^2-1)$ points, attaining the optimal quadratic order in the dimension. Assuming Zauner's conjecture, this bound improves to $7.5(d-1)^2$. For every odd $d \geq 9$, we also construct a spherical $7$-design with $110M$ points from a complex projective $3$-design with $M$ points in $\mathbb{CP}^{(d-3)/2}$.

0.21

Symmetry without a manifold: intrinsic dimension on orbits

cs.LG stat.ML ·2026-09-15 ·Chon-Fai Kam, Miloud Bessafi, Frédéric Cadet
unusual similarity 0.44
Abstract · 33 pages, 3 figures. Under review at NeurReps 2026 (NeurIPS workshop, Proceedings Track)

The standard geometric derivation of neural scaling exponents takes the intrinsic dimension of a data manifold as its input. On modular addition in $\mathbb{Z}_p$ that derivation has no input. The exact algebraic solution is an orbit of $\mathbb{Z}_p$ acting by isometries. Transitivity alone makes the ratio statistic underlying the standard dimension estimator a point mass, so the estimator is undefined, and here the two nearest neighbour distances coincide exactly. Breaking the symmetry at scale $ε$ returns a number, but one that tracks $1/ε$ with no scale free plateau. We show that the failure is general, since on any finite orbit of a group acting by isometries the estimator reports the resolution at which the set is probed rather than a dimension. What replaces the power law is exponential in hidden width, $L(h)=L_\infty+A\exp(-c\,h^α)$, with $R^2$ between 0.982 and 0.995 against 0.857 to 0.906 for a power law admitting the same floor and fitted under the same protocol. Where the data supply is sufficient the rate belongs to the regulariser rather than to the group, since weight decay moves $c$ by a factor of 47 while group order moves it by 1.10, a residual below seed to seed resolution, for every fixed $α$ between 0.75 and 2. The critical width falls with group order rather than rising, against capacity counting that assigns a fixed number of neurons to each irreducible representation.

0.34