pith. sign in

arxiv: 2605.05202 · v1 · submitted 2026-05-06 · 💻 cs.IT · math.IT

Optimizing Bit-Labeling of Voronoi Constellations

Carilyn Rumrill , David Muzzey , Connor Davis , Stephen Mackes , Dan Chew This is my paper

Pith reviewed 2026-05-08 15:45 UTC · model grok-4.3

classification 💻 cs.IT math.IT
keywords bit labelingVoronoi constellationsD4 latticeE8 latticebit error ratelattice codingbasis matrices
0
0 comments X

The pith

A search over basis matrices finds bit labelings for D4 and E8 lattices that cut bit error rates by 0.1 to 0.5 dB.

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

The paper sets out to improve bit-to-symbol assignments for Voronoi constellations drawn from the D4 and E8 root lattices by treating the choice of basis matrix as a variable that permutes the integer labels on fixed lattice points. With the constellation and other coding elements held constant, the authors apply a new search procedure whose performance metric is bit error rate itself. They report that the best labelings found this way outperform the bases most often cited in prior work. A reader would care because the gains appear at a practical BER operating point without any increase in transmit power or change in the underlying signal set, suggesting a low-cost way to extract more reliability from lattice codes already used in communications.

Core claim

After defining a search method over basis matrices that permute the integer labeling of lattice points and adopting bit error rate as the direct optimization criterion, the authors locate improved mappings for the four-dimensional D4 and eight-dimensional E8 root lattices. When the lattice constellation is fixed, these mappings deliver 0.1 dB gain for D4 and 0.5 dB gain for E8 relative to standard bases at a bit error rate of 10^{-4}.

What carries the argument

Search over basis matrices that permute integer labelings of fixed lattice points, scored by bit error rate.

Load-bearing premise

The search procedure and bit-error-rate metric locate labelings whose gains are due solely to the mapping itself and are not artifacts of simulation parameters or post-search selection.

What would settle it

Independent Monte Carlo simulation of the reported optimized bases versus the literature-standard bases under identical AWGN channel conditions, modulation, and decoder to check whether the 0.1 dB and 0.5 dB gaps at BER 10^{-4} reappear.

Figures

Figures reproduced from arXiv: 2605.05202 by Carilyn Rumrill, Connor Davis, Dan Chew, David Muzzey, Stephen Mackes.

Figure 1
Figure 1. Figure 1: A2 r = 4 marked lattice generated by the standard basis. the set of nearest neighbors of ⃗l0 and let A denote the integer vectors of these neighbors A = β −1 (N0). (12) We know for a given lattice point ⃗lk, the nearest neighbors Nk of that lattice point can be found by ⃗nk = β(β −1 (lk) + A mod r). (13) since we have assumed the constellation is shaped around a scaled voronoi region. Every lattice point h… view at source ↗
Figure 3
Figure 3. Figure 3: Hamming descent diagram: simplified Cayley graph continually view at source ↗
Figure 7
Figure 7. Figure 7: G8(8) bit error rate curves where the standard basis has HD = 9.13, the best unimodular gave HD = 3.86, and the worst unimodular gave HD = 10.2. VI. CONCLUSION Using a Hamming descent search, with HD as a perfor￾mance metric, we have found BER gain over the standard bases found in literature. We provided the theoretical lower bound for HD on the D4 and E8 lattice and have shown improvement towards that bou… view at source ↗
Figure 5
Figure 5. Figure 5: G4(8) bit error rate curves where the standard basis has HD = 2.33, the best unimodular gave HD = 2, and the worst unimodular gave HD = 3.41 view at source ↗
Figure 6
Figure 6. Figure 6: G8(4) bit error rate curves where the standard basis has HD = 7.35, the best unimodular gave HD = 3.86, and the worst unimodular gave HD = 7.91 view at source ↗
read the original abstract

We define a novel search method and performance metric as a technique for optimizing the bit-to-symbol map of the $D_4$ and $E_8$ root lattices in reference to bit error rate. We hold other sources of lattice gain constant by fixing the lattice constellation, and consider basis matrices that permute the integer labelings of the lattice points. After searching the possible basis matrices for $D_4$ and $E_8$, we found 0.1 dB of gain in $D_4$ bit error rate curves, and 0.5 dB of gain in $E_8$ compared to the standard bases commonly used in literature at a BER of $10^{-4}$.

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.

Referee Report

2 major / 1 minor

Summary. The manuscript defines a novel search method and performance metric for optimizing the bit-to-symbol mapping of Voronoi constellations drawn from the D4 and E8 root lattices. By holding the lattice geometry fixed and varying only the basis matrices that permute integer labelings of the points, the authors report discovering labelings that improve BER by 0.1 dB for D4 and 0.5 dB for E8 relative to standard bases at 10^{-4} BER.

Significance. If the gains prove reproducible, the separation of labeling optimization from lattice geometry offers a lightweight route to incremental coding gains in high-dimensional modulation without redesigning the constellation or decoder. The use of basis-matrix permutations to enumerate labelings is a structured approach that could generalize to other root lattices.

major comments (2)
  1. [Abstract] Abstract: the reported 0.1 dB (D4) and 0.5 dB (E8) gains at BER = 10^{-4} are presented without any information on the cardinality of the basis-matrix search space, the stopping criteria, the number of Monte-Carlo trials, or statistical significance testing. These omissions prevent verification that the improvements are robust rather than artifacts of an incomplete search or simulation variance.
  2. [Method] Method section (implied by the abstract's description of the 'novel search method and performance metric'): no validation is supplied showing that the performance metric used to rank candidate labelings correlates with actual BER curves across different SNR grids or decoder implementations. Without such cross-validation, it is unclear whether optimizing the metric necessarily produces the claimed BER improvement.
minor comments (1)
  1. [Abstract] The abstract would be clearer if it briefly stated the size of the enumerated basis set or the computational cost of the search.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and have made revisions to improve the manuscript's clarity and reproducibility.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the reported 0.1 dB (D4) and 0.5 dB (E8) gains at BER = 10^{-4} are presented without any information on the cardinality of the basis-matrix search space, the stopping criteria, the number of Monte-Carlo trials, or statistical significance testing. These omissions prevent verification that the improvements are robust rather than artifacts of an incomplete search or simulation variance.

    Authors: We agree that these details should have been included for full reproducibility. In the revised manuscript we have expanded both the abstract and the methods section to report the cardinality of the enumerated basis-matrix search space, the stopping criterion (exhaustive search over all admissible bases), the number of Monte-Carlo trials per BER point, and the statistical tests used to confirm that the observed gains exceed simulation variance. These additions directly address the concern that the reported improvements might be artifacts. revision: yes

  2. Referee: [Method] Method section (implied by the abstract's description of the 'novel search method and performance metric'): no validation is supplied showing that the performance metric used to rank candidate labelings correlates with actual BER curves across different SNR grids or decoder implementations. Without such cross-validation, it is unclear whether optimizing the metric necessarily produces the claimed BER improvement.

    Authors: The performance metric was constructed as a closed-form approximation to bit-wise error probability under the given labeling and lattice geometry. While the original manuscript relied on the final BER curves to demonstrate improvement, we acknowledge that an explicit correlation study was absent. In the revision we have added a validation subsection that compares metric rankings against simulated BER for a representative sample of labelings across multiple SNR points and for both maximum-likelihood and reduced-complexity decoders. The results show strong monotonic correlation, supporting that optimization of the metric produces the observed BER gains. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical search results over basis matrices are independent of inputs

full rationale

The paper defines a search procedure and metric, then reports measured BER gains from simulation on D4/E8 lattices after enumerating basis matrices. No equations, derivations, or self-citations are shown that reduce the reported 0.1 dB / 0.5 dB gains to fitted parameters, self-definitions, or prior author results by construction. The outcome is presented as an external empirical finding from the search, not a tautological renaming or forced prediction. This is the normal case of a self-contained experimental claim.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract introduces no explicit free parameters, mathematical axioms, or new postulated entities; the contribution is an empirical search over a discrete set of basis matrices with an unreported performance metric.

pith-pipeline@v0.9.0 · 5420 in / 1310 out tokens · 68293 ms · 2026-05-08T15:45:25.685794+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

  1. [1]

    2013 , publisher=

    Sphere packings, lattices and groups , author=. 2013 , publisher=

    work page 2013

  2. [2]

    IEEE Transactions on Information Theory , volume=

    A fast encoding method for lattice codes and quantizers , author=. IEEE Transactions on Information Theory , volume=. 1983 , publisher=

    work page 1983

  3. [3]

    IEEE Transactions on information theory , volume=

    Asymptotically optimal block quantization , author=. IEEE Transactions on information theory , volume=. 1979 , publisher=

    work page 1979

  4. [4]

    Canadian Journal of Mathematics , volume=

    Sphere packings and error-correcting codes , author=. Canadian Journal of Mathematics , volume=. 1971 , publisher=

    work page 1971

  5. [5]

    2014 , publisher=

    Lattice coding for signals and networks: A structured coding approach to quantization, modulation, and multiuser information theory , author=. 2014 , publisher=

    work page 2014

  6. [6]

    Journal of Lightwave Technology , volume=

    Low-complexity geometric shaping , author=. Journal of Lightwave Technology , volume=. 2020 , publisher=

    work page 2020

  7. [7]

    2025 , school=

    Design and analysis of lattice-based communication schemes: a study of constructions A and D , author=. 2025 , school=

    work page 2025

  8. [8]

    2021 IEEE International Symposium on Information Theory (ISIT) , pages=

    Designing Voronoi constellations to minimize bit error rate , author=. 2021 IEEE International Symposium on Information Theory (ISIT) , pages=. 2021 , organization=

    work page 2021

  9. [9]

    Dummit and Richard M

    David S. Dummit and Richard M. Foote , title =. 2003 , address =

    work page 2003

  10. [10]

    IEEE Transactions on Information Theory , volume=

    An algebraic approach to physical-layer network coding , author=. IEEE Transactions on Information Theory , volume=. 2013 , publisher=

    work page 2013

  11. [11]

    IEEE Transactions on Information Theory , volume=

    Encoding and indexing of lattice codes , author=. IEEE Transactions on Information Theory , volume=. 2018 , publisher=

    work page 2018

  12. [12]

    2005 , publisher=

    A First Course in Modular Forms , author=. 2005 , publisher=

    work page 2005