Pith. sign in

REVIEW 6 minor 39 references

Low-Pathwidth GRAND: Exact Likelihood-Ordered Enumeration for BPSK Transmission over Correlated Gaussian Noise

T0 review · 0 major / 6 minor · reviewed 2026-07-31 · grok-4.5

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

Pith's one-line read Exact maximum-likelihood GRAND is possible for BPSK in correlated Gaussian noise when the precision matrix has low pathwidth.

desk verdict Careful, well-validated construction that makes matched soft GRAND exact for sparse/low-pathwidth Gaussian precision—solid methods paper, moderate novelty, worth refereeing. read the letter →

arxiv 2607.28363 v1 pith:2AC6V2IP submitted 2026-07-30 cs.IT math.IT

Behrooz Razeghi This is my paper
classification cs.ITmath.IT MSC 94B3594A24
keywords GRANDmaximum-likelihooddecodingcorrelatedGaussiannoisesparseprecisionmatricespathwidthBPSKsoft-inputquadraticpseudo-Booleanenergy
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

Soft-input GRAND is only guaranteed to return a maximum-likelihood codeword if it tests noise-effect patterns in true conditional-likelihood order. With correlated Gaussian noise that order is not additive: the matched metric couples coordinates through the precision matrix, so rank-based and block-independent surrogates can miss the ML word. This paper shows that the candidate-dependent part of the Gaussian negative log-likelihood is exactly an observation-dependent quadratic energy on binary patterns, with edges only where the precision matrix is nonzero. When that interaction graph has small half-bandwidth or pathwidth, a trellis or bag-assignment layered graph with few states per layer lets suffix dynamic programming and best-first search emit every pattern in nondecreasing energy. With a full search and no early stop, the first codebook hit is therefore an ML codeword for any nonempty equiprobable binary codebook. On small codes the method matched exhaustive ML in every tested frame, and on longer codes it beat the usual block approximations in block-error rate.

What carries the argument

The observation-dependent quadratic energy Wr(z) together with a validated low-width path decomposition (or banded trellis) of its precision-induced interaction graph; suffix costs plus best-first complete-path enumeration produce the exact nondecreasing-energy order GRAND needs.

What would settle it

On any fixed short code and known Q, run exhaustive minimization of the matched Gaussian energy over the codebook and check whether LP-GRAND’s first codebook hit always lands in that minimizer set; disagreement on even one frame would refute the finite-block ML claim under the implemented metric.

Watch

Extended reading notes

Core claim

For BPSK in Gaussian noise with precision matrix Q, the candidate-dependent matched negative log-likelihood equals, up to a pattern-independent constant, a quadratic pseudo-Boolean energy Wr(z) whose pairwise terms sit exactly on the nonzero off-diagonal entries of Q. If that interaction graph admits a path decomposition of width w (or Q has half-bandwidth ν), complete paths in a layered graph with at most 2^{w+1} (resp. 2^ν) states per layer can be enumerated in nondecreasing Wr by suffix dynamic programming plus best-first search. Complete enumeration without abandonment then makes the first codebook hit an ML codeword.

Load-bearing premise

The exact ordered search stays practical only when the noise precision matrix induces an interaction graph of small pathwidth or half-bandwidth; the ML guarantee itself also needs a complete search with no early abandonment.

Editorial extensions

If this is right

  • Receivers that keep physical symbol order can still run likelihood-ordered GRAND when the noise precision is banded or low-pathwidth, without relying on interleaving or block-independence.
  • Additive reliability ranks and across-block product approximations are unnecessary for the ML query order whenever a validated low-width decomposition of Q exists.
  • The same layered graph that orders candidates also computes the full-space partition function and residual mass used for soft-output and abandonment analysis.
  • If Q is sparsified or mismatched, the enumerator remains exact for the modified metric, so order-stability bounds become the right tool for deciding how much mismatch is tolerable.

Reading between the lines

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

  • Channel estimation and equalizer design could treat pathwidth of the estimated precision as a first-class complexity budget for the decoder, not only MSE.
  • The method suggests a clean split: keep any codebook membership test, and put all correlation handling into a precision-driven path enumerator reusable across codes.
  • High-pathwidth residual interference would force either approximation, reordering/interleaving, or a different factorization—making pathwidth a diagnostic for when block-AI-style shortcuts become inevitable.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 6 minor

Summary. The paper develops LP-GRAND, a soft-input GRAND decoder for BPSK over Gaussian noise with a known sparse precision matrix Q. It shows that the candidate-dependent part of the matched negative log-likelihood is an observation-dependent quadratic pseudo-Boolean energy Wr(z) whose interaction graph is the off-diagonal sparsity pattern of Q. When that graph has half-bandwidth ν or admits a path decomposition of width w, suffix DP plus best-first complete-path enumeration produces noise-effect patterns in nondecreasing Wr. Under complete enumeration, fixed tie-breaking, and no abandonment, the first codebook hit is ML for any nonempty equiprobable binary codebook (Theorem IV.1). The construction is validated against exhaustive integer cost order, independent same-metric enumerators, and exhaustive codeword ML on two [20,12] codes (10,000 frames), and is compared empirically to block-product and ORBGRAND-AI baselines on [64,52] codes under first-order Gauss–Markov noise.

Significance. The work cleanly closes a gap between memoryless soft GRAND / ORBGRAND and matched decoding under correlated Gaussian noise when the precision graph has low pathwidth. The algebraic identity Er(z)=Er(0)+Wr(z) (Prop. III.1), the trellis and path-decomposition exactness theorems (V.2, V.6), and the standard GRAND ML ordering argument (IV.1) are proved in full in Appendix E. Strengths include: explicit real-arithmetic and complete-search hypotheses; stored-coefficient order checks against exact integer sorting and three independent generators; end-to-end agreement with exhaustive ML on 10k frames; public code; and carefully scoped comparisons (including metric equivalence with finite-block ISI when the NLL coincides). The pathwidth precondition is stated as a complexity requirement, not hidden in the ML claim. If the results hold as written, LP-GRAND is a useful exact reference and a practical decoder for banded or low-pathwidth precision models (e.g., AR(1), short ISI after whitening).

minor comments (6)
  1. [§VI-A, Fig. 3] §VI-A and (29): The nominal Eb/N0 parameterization fixes marginal variance while ρ varies, so the correlation sweep is not iso-innovation or iso-spectrum. This is stated, but a one-sentence reminder in the Fig. 3 caption would reduce misreading of the ρ-axis as an SNR-matched comparison.
  2. [§VI-D, Table IV] Table IV / Fig. 1 (right): Wall-clock times are correctly labeled as reference Python measurements. Consider moving the timing panel to an appendix or shortening it in the main text so that algorithmic counters (heap removals, retained paths) remain the primary complexity evidence.
  3. [Appendix G] Appendix G: The log-linear BLER=10^{-3} interpolation rests on 3–9 errors at 4 dB. The text already calls it exploratory; a single cautionary sentence in the main §VI-E or omitting the dB-gap column from any main-text summary would avoid over-citation of those numbers.
  4. [§V-B] §V-B: The four vertex-ordering heuristics are deterministic and the non-optimality caveat is clear. A brief note on whether RCM was always best on the path/ladder instances (as the median widths suggest) would help practitioners choose a default.
  5. [§II] Notation: y is both the hard-decision word and, in places, written y(r); s and xz switch between column-vector and Hadamard forms. A short notation paragraph early in §II would help.
  6. [Appendix F] Related work: SGRAND-ISI (arXiv:2603.08325) and the finite-block metric-equivalence discussion in Appendix F are careful. Ensure the published citation keys are updated if those preprints are accepted before camera-ready.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ML guarantee follows from matched metric plus standard GRAND ordering, with algebraic and exhaustive validation.

full rationale

The load-bearing chain is self-contained and non-circular. Proposition III.1 algebraically expands the Gaussian quadratic Er(z) into Wr(z)=Er(z)−Er(0) with explicit unary/pairwise coefficients from Q and r; this is identity, not a fit. Theorem IV.1 then applies the classical GRAND fact that nondecreasing matched negative log-likelihood order (complete enumeration, fixed ties, no abandonment) yields an ML codeword for equiprobable codewords—the paper states the complete-search and real-arithmetic hypotheses explicitly and does not smuggle the conclusion into the definition of Wr. Trellis and path-decomposition constructions (Theorems V.2, V.6) are standard layered representations that preserve path cost equal to Wr; enumeration correctness is checked against exhaustive integer sorting of stored coefficients and independent generators (state-list K-best, burst reference), not against a self-defined target. Auxiliary first-hit formulas are labeled occupancy-model estimates and calibrated separately. Empirical BLER comparisons are against external baselines on shared frames. No fitted-input-as-prediction, load-bearing self-citation uniqueness, or renaming of a known result as a first-principles derivation appears in the central claim.

Assumptions & free parameters 2 free parameters · 6 assumptions · 1 invented entities

The result rests on standard Gaussian BPSK channel math, the classical GRAND ML-ordering principle, and graph pathwidth/trellis dynamic programming. No fitted physical constants enter the ML theorem. Experimental knobs (ρ, rate, qmax, block size b) affect reported BLER, not the exact-order theorem. The main modeling commitments are known symmetric positive-definite Q, equiprobable codewords, and (for tractability) low pathwidth of G_Q.

free parameters (2)
  • Membership-query budget qmax = typically 20000 in main n=64 trials
    Finite abandonment budget used in BLER experiments (e.g. 20000); not part of the complete-enumeration ML theorem but controls practical error/abandonment tradeoffs.
  • Block length b for block baselines = b=8 principal; b=4 also reported
    Hand-chosen partition size for ExactBlockProduct and ORBGRAND-AI comparisons (b=4 or 8), not estimated from data but shapes baseline strength.
assumptions (6)
  • domain assumption Received model R=x(C)+N with N~N(0,Σ), Σ≻0, Q=Σ^{-1} known and used as the decoding metric.
    §II problem formulation; ML is w.r.t. this specified Gaussian model.
  • domain assumption Codewords equiprobable over nonempty C ⊆ {0,1}^n (ML=MAP).
    Theorem IV.1 and abstract ML guarantee.
  • domain assumption BPSK mapping x(c)=1-2c and hard-decision reference y from sign(r).
    §II; induces noise-effect parameterization z=y⊕c.
  • standard math Path decomposition / trellis dynamic programming correctly represents pairwise binary energies when bags cover vertices and edges with running intersection.
    Definition V.5, Theorem V.6; classical graphical-model fact.
  • domain assumption GRAND first-hit ML principle: nonincreasing likelihood query order + deterministic ties + no abandonment ⇒ first codebook member is ML.
    Cited from Duffy et al.; restated as Theorem IV.1 for Wr.
  • ad hoc to paper For efficient implementation, G_Q admits small half-bandwidth ν or pathwidth w under a constructible ordering.
    Central practical hypothesis of LP-GRAND; not true for arbitrary dense Q (Contributions 2, §V).
invented entities (1)
  • LP-GRAND enumerator (suffix DP + best-first complete-path search on precision path decomposition) independent evidence
    purpose: Emit all z in nondecreasing Wr order with codebook accessed only via membership test.
    Algorithmic construction rather than a new physical object; independent checks via exhaustive order and multi-generator agreement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Low-Pathwidth GRAND: Exact Likelihood-Ordered Enumeration for BPSK Transmission over Correlated Gaussian Noise." pith.science (2026). https://pith.science/paper/2AC6V2IP

@misc{pith2026260728363,
  author       = {Pith},
  title        = {Pith review of: Low-Pathwidth GRAND: Exact Likelihood-Ordered Enumeration for BPSK Transmission over Correlated Gaussian Noise},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2AC6V2IP}},
  note         = {Machine review of arXiv:2607.28363}
}
abstract

The finite-block maximum-likelihood (ML) guarantee of soft-input GRAND requires querying noise-effect patterns in nonincreasing conditional-likelihood order. Under correlated Gaussian noise, additive reliability metrics and independent-block approximations need not preserve this order because the matched metric contains cross-coordinate interactions; the first codebook hit need not induce an ML codeword. We develop Low-Pathwidth GRAND (LP-GRAND) for binary phase-shift keying (BPSK) with precision matrix $Q$. The candidate-dependent part of the Gaussian negative log-likelihood is an observation-dependent quadratic pseudo-Boolean energy whose interaction graph has edge $\{i,j\}$ exactly when $Q_{ij}\neq0$. If $Q$ has half-bandwidth at most $\nu$, this energy admits a trellis with at most $2^\nu$ states per layer; a path decomposition of width $w$ yields at most $2^{w+1}$ bag assignments per layer. In real arithmetic, suffix dynamic programming and best-first complete-path enumeration enumerate patterns in nondecreasing energy. With complete enumeration and no abandonment, the first codebook hit induces an ML codeword for any nonempty binary codebook with equiprobable codewords. LP-GRAND agreed with exhaustive codeword ML in all $10{,}000$ frames for two $[20,12]$ codes. At nominal $E_b/N_0=2$ dB, its empirical BLER was lower than that of each block-based approximation for six $[64,52]$ codes.

Figures

Figures reproduced from arXiv: 2607.28363 by the authors.

Figure 1
Figure 1. Left: induced frontier width under the coordinate and reverse Cuthill–McKee orderings for randomly permuted path, binary-tree, and ladder interaction [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Random-linear-code ensemble results as a function of code rate with [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Left: empirical BLER as a function of the first-order Gauss–Markov correlation coefficient at nominal [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Performance over the systematic [64, 52] random-linear-code ensemble with ρ = 0.5, qmax = 20,000, and 104 frames per nominal SNR. Left: BLER with 95% Wilson score intervals. Center: sample mean number of valid codebook-membership tests. Right: sample 99th percentile of…
Figure 5
Figure 5. Figure 5: Ordering sensitivity for K = 128. Left: top-128 overlap and union-restricted inversion rate as functions of the assumed decoder correlation ρb, using 100 common frames generated with ρ = 0.5 at nominal (Eb/N0)dB = 2.5. Markers denote sample means over frames, error bar…
Figure 6
Figure 6. Figure 6: Calibration of the random-codebook first-hit error estimate for [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Additional decoding and ordering results. Top: codeword-return rate [PITH_FULL_IMAGE:figures/full_fig_p027_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 4 linked inside Pith

  1. [3]

    Near shannon limit error- correcting coding and decoding: Turbo-codes. 1,

    C. Berrou, A. Glavieux, and P. Thitimajshima, “Near shannon limit error- correcting coding and decoding: Turbo-codes. 1,” inIEEE International Conference on Communications, vol. 2. IEEE, 1993, pp. 1064–1070. 12 0 0.25 0.5 0.75 0.9 Gauss–Markov correlation coefficient ½ 10−4 10−3 10−2 10−1 BLER LP-GRAND (matched) ExactBlockProduct, b = 8 ExactBlockProduct,...

    1993

  2. [9]

    Guessing random additive noise decoding with symbol reliability information (SRGRAND),

    K. R. Duffy, M. Médard, and W. An, “Guessing random additive noise decoding with symbol reliability information (SRGRAND),”IEEE Transactions on Communications, vol. 70, no. 1, pp. 3–18, 2021

    2021

  3. [14]

    The vertex separation number of a graph equals its path-width,

    N. G. Kinnersley, “The vertex separation number of a graph equals its path-width,”Information Processing Letters, vol. 42, no. 6, pp. 345–350, 1992

    1992

  4. [15]

    A linear time implementation of the reverse cuthill-mckee algorithm,

    W.-M. Chan and A. George, “A linear time implementation of the reverse cuthill-mckee algorithm,”BIT Numerical Mathematics, vol. 20, no. 1, pp. 8–14, 1980

    1980

  5. [16]

    The evolution of the minimum degree ordering algorithm,

    A. George and J. W. Liu, “The evolution of the minimum degree ordering algorithm,”Siam review, vol. 31, no. 1, pp. 1–19, 1989

    1989

  6. [17]

    A graph-theoretic study of the numerical solution of sparse positive definite systems of linear equations,

    D. J. Rose, “A graph-theoretic study of the numerical solution of sparse positive definite systems of linear equations,” inGraph theory and computing. Elsevier, 1972, pp. 183–217

    1972

  7. [18]

    Triangulation of graphs-algorithms giving small total state space,

    U. B. Kjærulff, “Triangulation of graphs-algorithms giving small total state space,” 1990

    1990

  8. [19]

    Probable inference, the law of succession, and statistical inference,

    E. B. Wilson, “Probable inference, the law of succession, and statistical inference,”Journal of the American Statistical Association, vol. 22, no. 158, pp. 209–212, 1927

    1927

Show all 39 references
  1. [20]

    List Viterbi decoding algorithms with applications,

    N. Seshadri and C.-E. W. Sundberg, “List Viterbi decoding algorithms with applications,”IEEE Transactions on Communications, vol. 42, no. 2/3/4, pp. 313–323, February 1994. 13 APPENDIXA ADDITIONALGAUSSIANNOISE-EFFECTMETRICRESULTS A.1. Stationary First-Order Gauss–Markov Noise ...

    1994

  2. [21]

    If Q is banded, its upper-triangular Cholesky factor can be chosen with the same half-bandwidth

    Thus, defining ¯r=Ar and v=An gives the equivalent transformed model ¯r=Ax+v , v∼ N(0, In). If Q is banded, its upper-triangular Cholesky factor can be chosen with the same half-bandwidth. The resulting whitening transform is banded and triangular but need not be Toeplitz. A b...

  3. [22]

    Soft maximum likelihood decoding using GRAND,

    A. Solomon, K. R. Duffy, and M. Médard, “Soft maximum likelihood decoding using GRAND,” inICC 2020-2020 IEEE International Con- ference on Communications (ICC). IEEE, 2020, pp. 1–6

    2020

  4. [23]

    Ordered reliability bits guessing random additive noise decoding,

    K. R. Duffy, W. An, and M. Médard, “Ordered reliability bits guessing random additive noise decoding,”IEEE Transactions on Signal Process- ing, vol. 70, pp. 4528–4542, 2022

    2022

  5. [24]

    Soft-output (so) grand and iterative decoding to outperform LDPC codes,

    P. Yuan, M. Médard, K. Galligan, and K. R. Duffy, “Soft-output (so) grand and iterative decoding to outperform LDPC codes,”IEEE Transactions on Wireless Communications, vol. 24, no. 4, pp. 3386–3399, 2025

    2025

  6. [25]

    Soft-output guessing codeword decoding,

    K. R. Duffy, P. Yuan, J. Griffin, and M. Médard, “Soft-output guessing codeword decoding,”IEEE Communications Letters, 2024

    2024

  7. [26]

    Using channel correlation to improve decoding-ORBGRAND-AI,

    K. R. Duffy, M. Grundei, and M. Médard, “Using channel correlation to improve decoding-ORBGRAND-AI,” inGLOBECOM 2023-2023 IEEE Global Communications Conference. IEEE, 2023, pp. 3585–3590

    2023

  8. [27]

    Verification of forecasts expressed in terms of probability,

    W. B. Glennet al., “Verification of forecasts expressed in terms of probability,”Monthly weather review, vol. 78, no. 1, pp. 1–3, 1950

    1950

  9. [28]

    On calibration of modern neural networks,

    C. Guo, G. Pleiss, Y . Sun, and K. Q. Weinberger, “On calibration of modern neural networks,” inInternational conference on machine learning. PMLR, 2017, pp. 1321–1330

    2017

  10. [29]

    T. A. Davis,Direct methods for sparse linear systems. SIAM, 2006

    2006

  11. [30]

    Low-density parity-check codes,

    R. Gallager, “Low-density parity-check codes,”IRE Transactions on information theory, vol. 8, no. 1, pp. 21–28, 1962

    1962

  12. [31]

    Factor graphs and the sum-product algorithm,

    F. R. Kschischang, B. J. Frey, and H.-A. Loeliger, “Factor graphs and the sum-product algorithm,”IEEE Transactions on Information Theory, vol. 47, no. 2, pp. 498–519, 2001

    2001

  13. [32]

    Near shannon limit error- correcting coding and decoding: Turbo-codes. 1,

    C. Berrou, A. Glavieux, and P. Thitimajshima, “Near shannon limit error- correcting coding and decoding: Turbo-codes. 1,” inIEEE International Conference on Communications, vol. 2. IEEE, 1993, pp. 1064–1070

    1993

  14. [33]

    Channel polarization: A method for constructing capacity- achieving codes for symmetric binary-input memoryless channels,

    E. Arikan, “Channel polarization: A method for constructing capacity- achieving codes for symmetric binary-input memoryless channels,”IEEE Transactions on information Theory, vol. 55, no. 7, pp. 3051–3073, 2009

    2009

  15. [34]

    Crc-aided decoding of polar codes,

    K. Niu and K. Chen, “Crc-aided decoding of polar codes,”IEEE communications letters, vol. 16, no. 10, pp. 1668–1671, 2012

    2012

  16. [35]

    List decoding of polar codes,

    I. Tal and A. Vardy, “List decoding of polar codes,”IEEE transactions on information theory, vol. 61, no. 5, pp. 2213–2226, 2015

    2015

  17. [36]

    Soft-decision decoding of linear block codes based on ordered statistics,

    M. P. Fossorier and S. Lin, “Soft-decision decoding of linear block codes based on ordered statistics,”IEEE Transactions on information Theory, vol. 41, no. 5, pp. 1379–1396, 1995

    1995

  18. [37]

    The guesswork of ordered statistics decoding: Guesswork complexity and decoder design,

    C. Yue, C. She, B. Vucetic, and Y . Li, “The guesswork of ordered statistics decoding: Guesswork complexity and decoder design,”IEEE Transactions on Information Theory, 2025

    2025

  19. [38]

    Capacity-achieving guessing random additive noise decoding,

    K. R. Duffy, J. Li, and M. Médard, “Capacity-achieving guessing random additive noise decoding,”IEEE Transactions on Information Theory, vol. 65, no. 7, pp. 4023–4040, 2019

    2019

  20. [39]

    Multi-code multi-rate universal maximum likelihood decoder using GRAND,

    A. Riaz, V . Bansal, A. Solomon, W. An, Q. Liu, K. Galligan, K. R. Duffy, M. Medard, and R. T. Yazicigil, “Multi-code multi-rate universal maximum likelihood decoder using GRAND,” inESSCIRC 2021-IEEE 47th European Solid State Circuits Conference (ESSCIRC). IEEE, 2021, pp. 239–246

    2021

  21. [40]

    A sub-0.8-pj/bit universal soft-detection decoder using ORBGRAND,

    A. Riaz, A. Yasar, F. Ercan, W. An, J. Ngo, K. Galligan, M. Médard, K. R. Duffy, and R. T. Yazicigil, “A sub-0.8-pj/bit universal soft-detection decoder using ORBGRAND,”IEEE Journal of Solid-State Circuits, vol. 60, no. 7, pp. 2645–2659, 2024

    2024

  22. [41]

    A finite-blocklength analysis for ORBGRAND,

    Z. Li and W. Zhang, “A finite-blocklength analysis for ORBGRAND,” arXiv preprint arXiv:2603.07526, 2026

    2026

  23. [42]

    Keep the bursts and ditch the interleavers,

    W. An, M. Médard, and K. R. Duffy, “Keep the bursts and ditch the interleavers,”IEEE Transactions on Communications, vol. 70, 2022

    2022

  24. [43]

    Decoding in the presence of ISI without interleaving orbgrand ai,

    K. R. Duffy, M. Grundei, J. A. Millward, M. Rangaswamy, and M. Medard, “Decoding in the presence of ISI without interleaving orbgrand ai,”arXiv preprint arXiv:2510.14939, 2025

    2025 arXiv

  25. [44]

    On universal decoding over discrete additive channels by noise guessing,

    H. K. Miyamoto and S. Yang, “On universal decoding over discrete additive channels by noise guessing,”arXiv preprint arXiv:2501.12971, 2025

    2025 arXiv

  26. [45]

    Parallelism empowered guessing random additive noise decoding,

    L. Wan, H. Yin, and W. Zhang, “Parallelism empowered guessing random additive noise decoding,”arXiv preprint arXiv:2510.01813, 2025

    2025 arXiv

  27. [46]

    GRAND for gaussian intersymbol interference channels,

    Z. Li and W. Zhang, “GRAND for gaussian intersymbol interference channels,”arXiv preprint arXiv:2603.08325, 2026

    2026

  28. [47]

    List Viterbi decoding algorithms with applications,

    N. Seshadri and C.-E. W. Sundberg, “List Viterbi decoding algorithms with applications,”IEEE Transactions on Communications, vol. 42, no. 2/3/4, pp. 313–323, February 1994

    1994

  29. [48]

    Finding the k shortest paths,

    D. Eppstein, “Finding the k shortest paths,”SIAM Journal on Computing, vol. 28, no. 2, pp. 652–673, 1998

    1998

  30. [49]

    Performance analysis and optimal design of ORB-type GRAND algorithms,

    L. Wan and W. Zhang, “Performance analysis and optimal design of ORB-type GRAND algorithms,”arXiv preprint arXiv:2606.28030, 2026

    2026 arXiv

  31. [50]

    A formal basis for the heuristic determination of minimum cost paths,

    P. E. Hart, N. J. Nilsson, and B. Raphael, “A formal basis for the heuristic determination of minimum cost paths,”IEEE transactions on Systems Science and Cybernetics, vol. 4, no. 2, pp. 100–107, 1968

    1968

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.