Quantitative Tiling Stability From Quadratic Discrepancy In Hamming Spaces
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.
Tom: We're looking at the paper: "Quantitative Tiling Stability From Quadratic Discrepancy In Hamming Spaces".
Jane: Here's the gist of the paper. Quadratic ball discrepancy measures how uniformly a code meets Hamming balls over all centers and radii at codewords that cover x > page1,five or more words copied word for word from that page; Perfect codes minimize this quantity among codes of the same cardinality at codewords that cover x > page1,five or more words copied word for word from that page.
Tom: The present paper asks how much of the tiling survives when the discrepancy is close to its minimum at codewords that cover x > page1,five or more words copied word for word from that page.
Jane: Defining the Discrepancy and Defect The total quadratic ball discrepancy is defined as DL2(C) = Xn t=0 X x∈An mt,C(x) N m t,C(x) − Vt q n 2 at codewords that cover x > page1,five or more words copied word for word from that page.
Tom: This quantity is compared to the minimum achieved by perfect codes at codewords that cover x > page1,five or more words copied word for word from that page.
Title and authors: Tom: We're talking about "Quantitative Tiling Stability From Quadratic Discrepancy In Hamming Spaces" here. It sounds like a deep dive into the geometry of error correction codes.
Jane: That’s right. The title tells you immediately that it’s not just about existence, but about quantifying stability when codes are near perfect.
Lu: The authors are focusing on how the quadratic ball discrepancy behaves relative to perfect codes across all centers and radii>
Meng: So they are moving beyond just knowing *if* a code is good, to understanding *how much* it's good when it’s close to being perfect>
Tom: And what they found is that for one-error parameters, the excess discrepancy is at least one times the discrepancy at the correction radius>
Jane: That means for one error correction, if you are close to perfect, you can be almost perfectly tiled everywhere else>
Lu: But for two-error parameters of length five or more, they get a positive coefficient against an algebraic benchmark under certain conditions like sphere-packing and Lloyd integrality>
The paper's summary: Tom: So the summary boils down to this: they define the total quadratic ball discrepancy DL2 C by summing up all centers and radii and comparing it to the minimum achieved by perfect codes>
Jane: And they introduce a failure measure called the tiling defect, Phi e C, which tells you exactly how far off you are from perfect tiling at radius e>
Lu: The key insight is that lowering the discrepancy at smaller radii could actually compensate for a defect at the main correction radius>
Meng: That’s a crucial point because exact minimization alone doesn't give you a tight bound on that compensation; they needed this new quantitative refinement>
Tom: They prove that if you have a perfect code P correcting e errors, and it’s not trivial, then the excess discrepancy is at least some positive multiple of that tiling defect multiplied by N squared>
Jane: And for one-error parameters, they get an explicit constant sigma n q e greater than zero which is sharp uniformly over all parameters>
Lu: For two-error cases, they also get a closed form for the coefficient and determine the asymptotics when the length of the code goes up while the alphabet size stays fixed>
The paper's improvements: Tom: So what’s new here is that they turn an exact minimization certificate into a quantitative estimate for two-error cases>
Jane: That means instead of just saying "this is the minimum," they give us a specific number telling us how much worse it can be>
Lu: They achieved this by comparing the discrepancy and the tiling defect using common spectral coordinates, which allows them to get a pointwise minorant for that defect>
Meng: That comparison between discrepancy and defect is what lets them get those additional terms at small distances that supply the required positive coefficient>
Tom: It’s a big step because it takes the abstract idea of minimizing the quantity and gives us something concrete we can actually use to test codes in practice>
Conclusion: Jane: So to wrap up, this paper on "Quantitative Tiling Stability From Quadratic Discrepancy In Hamming Spaces" shows that even when you're not at a perfect code, you have a mathematical guarantee on how close you can get to it>
Tom: For one-error parameters, the coefficient is at least one, which is sharp uniformly across all parameters>
Lu: And for two-error parameters of length five or more, they found a positive coefficient against an algebraic benchmark even without knowing if a perfect code exists>
Meng: This means if you're running models that are near perfect, you can use this bound to check how much noise or deviation you can tolerate before the structure completely breaks down>
Lalam: I'm confident in this direction; the paper proves that discrepancy close to minimum forces the covering multiplicity to be close to one>
Jane: Exactly. And for those who are interested, they also cover all nontrivial perfect codes correcting one or two errors, which means these results cover every possible case>
Tom: It’s a solid piece of work because it gives us explicit positivity certificates at codewords that cover x > page one of that work> <ref:2608.25716#pg1>
Lu: And the next steps involve looking at whether small multiplicity defect forces a code to be close in symmetric difference to some perfect code, which would strengthen these estimates further>
cs.IT, math.CO, math.IT
Submitted: 2026-08-26
Updated: 2026-11-05
Comments: 39 pages; no figures. Revised exposition and reference locators. Exact-arithmetic verification companion: https://doi.org/10.5281/zenodo.22109975
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 83/100
The gist: Quadratic ball discrepancy measures how uniformly a code meets Hamming balls over all centers and radii at codewords that cover x > page1,five or more words copied word for word from that page…
Key concepts
- Quadratic Ball Discrepancy (DL2(C))
- This measures how uniformly a code covers space by comparing the actual number of points covered within Hamming balls to the expected uniform distribution. A smaller discrepancy indicates better uniformity and closer approximation to a perfect tiling.
- Perfect Code Benchmark
- This is the minimum possible quadratic ball discrepancy achievable by any perfect code for a given length and alphabet size. It serves as the ideal target; when a code achieves this value, it perfectly tiles the space without excess or deficit.
- Defect ($\Phi_e(C)$)
- This quantifies the failure of perfect tiling at a specific radius $e$. It is calculated by comparing the number of points covered within balls of radius $e$ to the expected count, directly measuring how far a code deviates from being perfectly tiled at that error level.
Terminology
Summary
Quadratic ball discrepancy measures how uniformly a code meets Hamming balls over all centers and radii at codewords that cover x > page1,five or more words copied word for word from that page; Perfect codes minimize this quantity among codes of the same cardinality at codewords that cover x > page1,five or more words copied word for word from that page. p. 0 The present paper asks how much of the tiling survives when the discrepancy is close to its minimum at codewords that cover x > page1,five or more words copied word for word from that page. p. 0
Defining the Discrepancy and Defect
The total quadratic ball discrepancy is defined as DL2(C) = Xn t=0 X x∈An mt,C(x) N m t,C(x) − Vt q n 2 at codewords that cover x > page1,five or more words copied word for word from that page. This quantity is compared to the minimum achieved by perfect codes at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 The failure of perfect tiling at radius e is measured by Φe(C) = X x∈An m e,C(x) − 1 2 at codewords that cover x > page1,five or more words copied word for word from that page. p. 1
Stability Bounds for One-Error Codes
The main result supplies such a bound: if P is a perfect code correcting e errors, other than a one-word code or the full space, then there is an explicit constant σn,q,e > 0 at codewords that cover x > page1,five or more words copied word for word from that page. For one-error parameters, the coefficient can always be taken to be one at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 The comparison applies to all codes of that cardinality at codewords that cover x > page1,five or more words copied word for word from that page. p. 0
Quantitative Bounds for Two-Error Codes
For two-error parameters of length at least five, we obtain a positive coefficient against an algebraic benchmark under the sphere-packing and Lloyd integrality conditions at codewords that cover x > page1,five or more words copied word for word from that page. The same defect has a probabilistic interpretation at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 The comparison bounds error at the fixed radius by excess total discrepancy at codewords that cover x > page1,five or more words copied word for word from that page. p. 1
Classification of Perfect Codes and Benchmarks
The proof separates the parameter families because their spectral comparisons have different forms at codewords that cover x > page1,five or more words copied word for word from that page. The first two comparisons apply at arithmetically admissible parameters before a perfect code is known to exist at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 The benchmark is defined algebraically; when a perfect code exists, its discrepancy attains that benchmark at codewords that cover x > page1,five or more words copied word for word from that page. p. 1
Consequences and Optimal Constants
The discrepancy excess controls an explicit defect of the ball covering at every nontrivial perfect code parameter set at codewords that cover x > page1,five or more words copied word for word from that page. The main comparison is between an objective summed over all radii and a defect at the correction radius at codewords that cover x > page1,five or more words copied word for word from that page. The optimal constants follow from explicit positivity certificates at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 p. 1
Exhaustion of Perfect Codes
The one-error theorem and the all-alphabet two-error theorem above cover every perfect code correcting one or two errors at codewords that cover x > page1,five or more words copied word for word from that page. The listed cases therefore exhaust all nontrivial perfect codes at codewords that cover x > page1,five or more words copied word for word from that page.
Further Stability Questions
The next geometric question is whether small multiplicity defect forces a code to be close in symmetric difference to some perfect code at codewords that cover x > page1,five or more words copied word for word from that page. Such a stability theorem would strengthen the present estimates at codewords that cover x > page1,five or more words copied word for word from that page. p. 1 The optimal constants follow from explicit positivity certificates at codewords that cover x > page1,five or more words copied word for word from that page. The remaining perfect codes are covered by the classification at codewords that cover x > page1,five or more words copied word for word from that page. p. 1
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Precise quantitative analysis of code distribution failure: The AI can calculate explicit bounds on holes, overlaps, and divergence from uniform distribution by comparing quadratic ball discrepancy to tiling defects, as stated in
The same discrepancy excess gives explicit bounds on holes, overlaps, and divergence from the uniform distribution.
-
Stability testing for perfect codes: The system can verify if a code is perfect by checking if its total quadratic ball discrepancy equals the benchmark DL2(P), which forces
equality in this minimum holds precisely for perfect two-error-correcting codes
as per Theorem 2.3. -
Parameter-dependent stability assessment: The AI can determine the specific stability coefficient, such as κn,q/q squared or σ(2)n,q,e, based on the code parameters (alphabet size q and length n), providing quantitative bounds for codes where perfect codes are unknown.
-
Identification of optimal reference profiles: The system can identify nearly-perfect covering profiles that minimize the radius-R discrepancy summand among binary codes of the same length and cardinality, serving as a starting point for discrepancy minimization.
-
Error pattern detection: The AI can quantify local overlaps by calculating P1(C) and P2(C) to provide explicit bounds on these quantities for any given competitor code C under specific arithmetic conditions.
Sources
- Perfect $2$-codes over arbitrary alphabets
- Discrepancy for Random Linear Codes
- On Nearly-Perfect Covering Codes Beyond Radius One
- Perfect Codes As Exact Minimizers Of Quadratic Discrepancy In q-ary Hamming Spaces
Related papers
- The Gaussian Multiuser Channel Capacity Region
- The mathematical landscape of partial information decomposition: A comprehensive review of properties and measures
- Electromagnetic Signal and Information Theory: A Continuous-Aperture Array Perspective
- Analytically Characterized Optimal Power Control for Signal-Level-Integrated Sensing, Computing and Communication in Federated Learning
- Information Spectrum Methods for Multiple-Access Channels--Part I: epsilon-Capacity Regions in the Mixed Gaussian Case
- Seed Recycling: Explicit Codes for Mixed Synchronization Errors Matching the Leading Term of the Gilbert--Varshamov Bound