Universal Decoding over Finite-State Additive Channels via Noise Guessing

Henrique K. Miyamoto; Sheng Yang

REVIEW 2 major objections 5 minor 2 cited by

Universal Decoding over Finite-State Additive Channels via Noise Guessing

T0 review · 2 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Noise guessing can match maximum-likelihood decoding on unknown finite-state additive channels.

desk verdict Solid theory paper extending universal GRAND to finite-state additive channels; the universality and complexity bounds hold up, but the empirical 'consistently outperform' claim overreaches because the comparison baseline runs at a higher rate and the figures lack error bars. read the letter →

arxiv 2501.12971 v2 pith:PSRUREK5 submitted 2025-01-22 cs.IT math.IT

Henrique K. Miyamoto , Sheng Yang This is my paper · ORCID

classification cs.ITmath.IT MSC 94A1594A40

keywords additivechannelserrorexponentfinite-statetypesguessworkuniversaldecodingGRANDKTestimatorrandomcoding

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

This paper asks whether a receiver can decode reliably over an additive channel whose noise statistics it does not know, without paying the usual estimate-then-decode penalty. It studies noise generated by a finite-state process with deterministic transitions (unifilar), and proposes two noise-guessing decoders: one that enumerates noise sequences in a fixed order, and one that draws them at random. The central claim is that both are random-coding universal: for long block lengths their error probability decays with the same exponent as maximum-likelihood decoding, even though the channel law is never learned. The paper also bounds the average number of guesses, showing the same complexity exponent as a noise-guessing decoder that knows the noise distribution exactly.

What carries the argument

The argument runs on the finite-state type decomposition $P_{\theta}(zzz)=2^{-n(H(\hat P_{zzz})+D(\hat P_{zzz}\|P_\theta))}$, which holds because the next-state function $f(\cdot,s)$ is injective for each state, so the state sequence is determined by the symbol sequence and empirical counts over (symbol, previous-state) pairs define a type. The maximising and weighting estimators — the normalised maximum-likelihood distribution and the KT mixture (a Bayesian average over Dirichlet priors) — stand in for the unknown law, and the type-counting lemmas give the rank bounds $G_m(zzz)\le 2^{n(H(\hat P_{zzz})+\zeta(n))}$ and $G_w(zzz)\le 2^{n(H(\hat P_{zzz})+\epsilon(n))}$ that turn guessing into decoding. The randomised version replaces the fixed ranking with sampling from these estimators, which preserves the error exponent while making each guess cheap to generate and parallelisable.

What would settle it

Take a binary additive channel whose noise is a hidden Markov process that is not unifilar (for instance, two states that can both emit 0 or 1, with a next-state rule not determined by the emitted symbol), and compare the empirical error exponent of the maximising guessing decoder against maximum-likelihood decoding at block lengths 100, 200, and 400. If the redundancy does not shrink like $O(\log n/n)$, or the decoder's exponent falls below the ML exponent, the unifilar injectivity assumption is doing essential work.

Watch

Extended reading notes

Core claim

The paper's central claim is that noise guessing can be made universal for all additive unifilar channels: a decoder that knows only that the noise belongs to this family, not its parameters, can match the random-coding error exponent of the maximum-likelihood decoder. Concretely, Theorem 1 bounds the redundancy of deterministic guessing with the maximising (NML) metric by $1/n+\zeta(n)+\delta(n)=O(\log n/n)$, uniformly over channels, and Theorem 3 gives the analogous $2/n+\zeta(n)+\delta(n)$ bound for randomised guessing; the weighting (KT) versions replace $\zeta(n)$ with $\epsilon(n)$. Theorems 2 and 4 bound the average guess count, showing both strategies achieve the complexity exponent $\min\{H_{1/2}(P_\theta), \log|\mathcal A|-R\}+O(\log n/n)$, the same asymptotic complexity as GRAND when the noise distribution is known. Numerical experiments on first-order Markov additive channels show that both universal decoders consistently beat a training-based strategy that reserves channel uses for parameter estimation.

Load-bearing premise

The whole argument rests on the noise process being unifilar with an injective next-state function, so the sequence of hidden states is fully recoverable from the observed noise symbols; if two different states can lead to the same next state for a given symbol, the type decomposition and all counting bounds collapse.

Editorial extensions

If this is right

A receiver on an unknown additive finite-state channel can decode at maximum-likelihood reliability with no training phase; the redundancy gap vanishes like $O(\log n/n)$.
Universality does not degrade the asymptotic guessing complexity: both proposed decoders have the same complexity exponent as GRAND with exactly known noise statistics.
The randomised decoder can be implemented with per-guess cost linear in block length and state count, and its guess generation can be parallelised without coordination, which matters when the noise process has many states.
The weighting (KT) version admits a sequential sampling rule, so universal decoding can be run in a streaming fashion without enumerating types.

Reading between the lines

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

An extension the paper does not pursue: the same type-based redundancy proof should carry over to channels with known deterministic side information, since only the unifilar structure is used; this is directly testable.
The complexity bounds suggest that on channels with large state spaces, randomised guessing should win in wall-clock time despite drawing more guesses, because deterministic guessing must enumerate all types; a benchmark on third-order Markov noise could check this.
The injectivity assumption is the natural stress point: if the noise were a hidden Markov source whose state is not recoverable from the symbol sequence, the type decomposition would need to be replaced by averaging over state paths, likely costing an extra logarithmic factor in redundancy.

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.

Desk Editor's Note

read the letter

Read this paper. The central claim is correct: deterministic and randomized noise-guessing decoders are random-coding universal for additive unifilar channels, with non-asymptotic redundancy O(log n/n) and complexity exponents that match true GRAND with known channel law. That is a real advance over previous memoryless-only results (Joudeh; the authors' own KT decoder). The randomized variant, which draws noise sequences sequentially with linear per-iteration generation and can be fully parallelized, is a useful contribution in its own right.

The proofs are what make this paper worth taking seriously. The finite-state type method is applied carefully; the bounds in Theorems 1-4 are derived in full; the complexity exponents are compared to the true H_{1/2}, not to a fitted value. The assumption of unifilarity with injective next-state map is explicit and honestly stated: if states are hidden, the type-counting machinery and the complexity bounds do not apply. That is a scope limitation, not a flaw. No circularity: the estimators (NML and KT) are fixed, parameter-free, and the theorems hold uniformly over the family.

The soft spot is the numerical section, and it is entirely fixable. The training-based baseline uses 8 of 63 channel uses for pilot and codes 55 data bits, so its rate is 55/63; the proposed schemes use a modified BCH code with effective rate 50/63. The baseline is operating at a higher information rate, so its worse BLER is expected. Comparing BLER without holding rate and energy fixed does not demonstrate superiority. In addition, the figure points have no error bars or confidence intervals, so the word 'consistently' in the abstract is not supported by the evidence. The Appendix A normalization slip (defining the circular extension with /n rather than /(n+c)) is minor; the algebra uses the correct normalization.

For a reader: this is an information theory paper for people working on universal decoding, GRAND, and low-latency short-packet communication. The theory is solid and the empirical overreach is contained in Section V. It deserves a serious referee; I expect the reviewers will ask for a fairer experimental comparison (matched rates, repeated trials) and possibly a clearer statement that the practical advantage is demonstrated only qualitatively. I would cite the main theorems and would bring it to a reading group. Send it to review.

Referee Report

2 major / 5 minor

Summary. The paper studies universal decoding for discrete additive channels whose noise is generated by a unifilar finite-state process. Two low-complexity decoders are proposed: deterministic noise guessing, which queries noise sequences in an order induced by the NML ('maximising') or KT ('weighting') score, and randomized noise guessing, which draws noise sequences from these distributions until a codeword is found. The main theoretical claims are Theorems 1 and 3, which upper-bound the redundancy against ML decoding by O(log n / n) uniformly over the channel family, and Theorems 2 and 4, which upper-bound the guessing complexity exponent by min{H_{1/2}(P_theta), log|A| - R} + O(log n / n). The proofs are based on finite-state types, a stochastic-decoder comparison lemma, and concentration of the number of distinct codewords. Section V reports simulations on binary order-1 Markov additive channels comparing the proposed decoders with matched GRAND, a training-based decoder, and a mismatched memoryless decoder.

Significance. If the proofs hold, this is a substantial contribution. It extends GRAND-style universal decoding from memoryless channels to a general unifilar finite-state class, gives explicit non-asymptotic redundancy and complexity bounds, and introduces a randomized variant that avoids type enumeration and is parallelizable. The estimators used are parameter-free, and the theorems give uniform guarantees over all channel parameters rather than point estimates. The paper's theoretical core is well supported: the type-method arguments are complete, and the main derivations in Sections III and IV check out. The empirical 'consistently outperform' claim is not yet supported, because the comparison in Section V is not rate-matched and the reported points lack uncertainty quantification; this affects the practical-advantage claim but not the universality or complexity theorems.

major comments (2)

[Section V, Fig. 1 and abstract] The numerical comparison is not rate-matched. The proposed scheme uses a modified BCH code with effective rate (k-1)/n = 50/63, while the training-based baseline spends 8 of the n=63 channel uses on a known training sequence and encodes 55 information bits, giving an effective rate of 55/63. At the same channel parameters p, the lower-rate code is expected to have a lower BLER even without any universality advantage, so Figs. 1(a)-(d) do not establish that the universal strategies 'consistently outperform' the training-based strategy as claimed in the abstract. The comparison should be repeated at equal information rate (including the training overhead), or the claim should be weakened to a rate-unmatched illustration.
[Section V, Figs. 1-2] The figures do not report how many trials or channel/code realisations each operating point is based on, and no error bars or confidence intervals are provided. Since the abstract and the conclusion use the word 'consistently', the empirical claim requires some quantification of variability; otherwise the statement should be limited to the specific runs shown.

minor comments (5)

[Appendix A] The displayed definition of the circular extension distribution has eP_zzz(z,s) := e a_zzz(z,s)/n; since the circular sequence has length n+c, the normalization should be /(n+c). The subsequent algebra in (57) uses the correct normalization, so the proof is not affected, but the displayed definition should be corrected.
[Theorems 2 and 4] The statements should explicitly assume R < log|A| (or treat R = log|A| separately). The proof of Theorem 4 requires E[K_M]/M - k*/M > 0, which uses 1/n - 2^{nR}/(2|A|^n) > 0 and therefore fails when R = log|A|.
[Section IV.B] The notation Q_{theta,tilde P_Z}(error) for the average number of guesses is confusing; it should be Q_{theta,tilde P_Z}(n), consistent with Theorems 2 and 4.
[Section V] Please clarify whether the training-based decoder also uses the antipodal-pair-reduced code or the original BCH code; the effective-rate calculation depends on this and is part of the rate-mismatch concern.
[Section II.C] In (8), the probability expression omits the initial-state dependence in the notation; this is stated earlier, but keeping s0 in the equation would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the universality and complexity proofs are derived from external type-counting, NML/KT estimation bounds, and Rényi-entropy variational results, with no fitted input renamed as a prediction.

full rationale

The central derivation is self-contained in the relevant sense: the paper proves the claimed redundancy and complexity bounds rather than importing them from its own definitions. Theorem 1 upper-bounds the redundancy of the maximising and weighting deterministic guessing decoders by comparing equivocation sets (Lemma 5 from Feder-Lapidoth, Lemma 6 proved via finite-state type classes) and bounding the guessing rank (Lemma 7 via the Shtarkov NML and KT distributions, together with Lemma 1 type-counting from Csiszár). The random-coding error exponent is compared with the true ML decoder P_theta,ML, not with a quantity fitted from the same data. Theorems 3 and 4 use the same external ingredients, plus Lemma 3 (Rényi entropy variational formula, sourced externally) and Lemma 4; the proof of Lemma 4 in Appendix A uses the standard circular-extension construction of Davisson-Longo-Sgarro, which is a technique name, not a circular argument. Complexity exponents are benchmarked against the true Rényi entropy H_{1/2}(P_theta) of the channel, and the additive terms zeta(n), delta(n), epsilon(n), sigma_theta(n) are explicit finite-n corrections, not fitted values. The only self-citation, [14], is used as an analogy for the memoryless case and for the antipodal-pair code modification; it is not load-bearing for the universality or complexity proofs. The experimental claim that the universal strategies outperform the training-based baseline may be questioned on fairness grounds (the baseline operates at a higher effective rate and the figures show no uncertainty), but that is an empirical-support issue, not circularity.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central results are derived from external information-theoretic tools (method of types, KT estimator bounds, Rényi entropy variational formula). The only domain assumptions are the unifilar/injective structure and, for complexity bounds, stationarity and irreducibility of the induced Markov chain. No free parameters are fitted to data.

assumptions (6)

standard math Method of types for finite-state sequences (Lemma 1, from [23])
Used to count type classes and bound the size of equivocation sets.
standard math Rényi entropy variational formula for irreducible Markov chains (Lemma 3, from [31])
Used to upper bound expected guessing ranks in Theorems 2 and 4.
domain assumption Unifilar additive channel with injective next-state map f(., s)
Ensures the state sequence is a function of the symbol sequence; invoked in Section II.B and throughout.
domain assumption For complexity results, the induced Markov chain is stationary and irreducible (family F*)
Required for the circular extension argument in Lemma 4 (Appendix A).
domain assumption Uniform random coding over A^n and equiprobable messages
Defines the random-coding universality notion used in all theorems.
standard math KT estimator bound relative to the type distribution (Lemma 2, from [30])
Used to show the weighting decoder is within O((log n)/n) of the maximising decoder.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Universal Decoding over Finite-State Additive Channels via Noise Guessing." pith.science (2026). https://pith.science/paper/PSRUREK5

@misc{pith2026250112971,
  author       = {Pith},
  title        = {Pith review of: Universal Decoding over Finite-State Additive Channels via Noise Guessing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PSRUREK5}},
  note         = {Machine review of arXiv:2501.12971}
}

read the original abstract

We study universal decoding over unknown discrete additive channels determined by a finite-state (unifilar) random process. Aiming at low-complexity decoders, we study variants of noise-guessing decoders that use estimators for the probability of a noise sequence when the actual channel law is unknown. A deterministic version produces noise sequences in a fixed order, and a new randomised version draws them at random, until finding a sequence that, subtracted from the received sequence, results in a valid codeword. We show that both strategies are random-coding universal (i.e. have the same random-coding error exponent as the optimal maximum likelihood decoding), and derive upper bounds for their complexity. Numerical examples in additive Markov channels illustrate the proposed methods' performance, showing that they consistently outperform a more usual training-based strategy.

Figures

Figures reproduced from arXiv: 2501.12971 by the authors.

**Figure 2.** Complexity of different decoders in order- [PITH_FULL_IMAGE:figures/full_fig_p025_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Low-Pathwidth GRAND: Exact Likelihood-Ordered Enumeration for BPSK Transmission over Correlated Gaussian Noise
cs.IT 2026-07 accept novelty 6.0 of 10

For BPSK in correlated Gaussian noise, low-pathwidth path decompositions of the precision graph enable exact likelihood-ordered GRAND enumeration and ML decoding under complete search.
Closed-Loop Bayesian Bandit Encoder with GRAND Receiver for a Bursty Interference Channel
cs.IT 2026-07 conditional novelty 6.0 of 10

In simulation, a receiver that learns bursty interference makes non-interleaved transmission preferable to interleaved, reversing the preference once the learned decoder is active.

Reference graph

Works this paper leans on

36 extracted references · 34 canonical work pages · cited by 2 Pith papers

[1]

Reliable communication under channel uncertainty,

A. Lapidoth and P. Narayan, “Reliable communication under channel uncertainty,” IEEE Trans. Inf. Theory, vol. 44, no. 6, pp. 2148–2177, 1998

work page 1998
[2]

Universal decoding for channels with memory,

M. Feder and A. Lapidoth, “Universal decoding for channels with memory,” IEEE Trans. Inf. Theory , vol. 44, no. 5, pp. 1726–1745, 1998

work page 1998
[3]

Nonprobabilistic mutual information with memory,

V . D. Goppa, “Nonprobabilistic mutual information with memory,” Problems Control Inf. Theory, vol. 4, pp. 97–102, 1975

work page 1975
[4]

Csiszár and J

I. Csiszár and J. Körner, Information Theory: Coding Theorems for Discrete Memoryless Systems , 2nd ed. Cambridge, UK: Cambridge Univ. Press, 2011

2011
[5]

Universal decoding for finite-state channels,

J. Ziv, “Universal decoding for finite-state channels,” IEEE Trans. Inf. Theory , vol. 31, no. 4, pp. 453–460, 1985

work page 1985
[6]

Universal decoding for memoryless Gaussian channels with a deterministic interference,

N. Merhav, “Universal decoding for memoryless Gaussian channels with a deterministic interference,” IEEE Trans. Inf. Theory, vol. 39, no. 4, pp. 1261–1269, 1993

work page 1993
[7]

On the universality of the LZ-based decoding algorithm,

A. Lapidoth and J. Ziv, “On the universality of the LZ-based decoding algorithm,” IEEE Trans. Inf. Theory, vol. 44, no. 5, pp. 1746–1755, 1998

work page 1998
[8]

Universal composite hypothesis testing: a competitive minimax approach,

M. Feder and N. Merhav, “Universal composite hypothesis testing: a competitive minimax approach,” IEEE Trans. Inf. Theory, vol. 48, no. 6, pp. 1504–1517, 2002

work page 2002

Show all 36 references

[9]

Universal decoding for frequency-selective fading channels,

O. Shayevitz and M. Feder, “Universal decoding for frequency-selective fading channels,” IEEE Trans. Inf. Theory, vol. 51, no. 8, pp. 2770–2790, 2005

2005
[10]

Universal decoding for arbitrary channels relative to a given class of decoding metrics,

N. Merhav, “Universal decoding for arbitrary channels relative to a given class of decoding metrics,” IEEE Trans. Inf. Theory, vol. 59, no. 9, pp. 5566–5576, 2013

2013
[11]

Universal decoding for Gaussian intersymbol interference channels,

W. Huleihel and N. Merhav, “Universal decoding for Gaussian intersymbol interference channels,” IEEE Trans. Inf. Theory, vol. 61, no. 4, pp. 1606–1618, 2015

2015
[12]

Exact random coding exponents and universal decoders for the asymmetric broadcast channel,

R. Averbuch and N. Merhav, “Exact random coding exponents and universal decoders for the asymmetric broadcast channel,” IEEE Trans. Inf. Theory , vol. 64, no. 7, pp. 5070–5086, 2018

2018
[13]

Universal decoding for the typical random code and for the expurgated code,

R. Tamir and N. Merhav, “Universal decoding for the typical random code and for the expurgated code,” IEEE Trans. Inf. Theory, vol. 68, no. 4, pp. 2156–2168, 2022

2022
[14]

On universal decoding over memoryless channels with the Krichevsky–Trofimov estimator,

H. K. Miyamoto and S. Yang, “On universal decoding over memoryless channels with the Krichevsky–Trofimov estimator,” in Proc. 2024 IEEE Int. Symp. Inf. Theory (ISIT) , 2024, pp. 1498–1503

2024
[15]

On the decoding of convolutional codes on an unknown channel,

A. Lapidoth and J. Ziv, “On the decoding of convolutional codes on an unknown channel,” IEEE Trans. Inf. Theory , vol. 45, no. 7, pp. 2321–2332, 1999. 32

1999
[16]

Capacity-achieving guessing random additive noise decoding,

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

2019
[17]

Keep the bursts and ditch the interleavers,

W. An, M. Médard, and K. R. Duffy, “Keep the bursts and ditch the interleavers,” IEEE Trans. Commun. , vol. 70, no. 6, pp. 3655–3667, 2022

2022
[18]

On guessing random additive noise decoding,

H. Joudeh, “On guessing random additive noise decoding,” in Proc. 2024 IEEE Int. Sympo.Inf. Theory (ISIT) , 2024, pp. 1291–1296

2024
[19]

Ensemble-tight second-order asymptotics and exponents for guessing-based decoding with abandonment,

V . Y . F. Tan and H. Joudeh, “Ensemble-tight second-order asymptotics and exponents for guessing-based decoding with abandonment,” 2025. [Online]. Available: https://arxiv.org/abs/2502.05959

2025 arXiv
[20]

Universal randomized guessing with application to asynchronous decentralized brute–force attacks,

N. Merhav and A. Cohen, “Universal randomized guessing with application to asynchronous decentralized brute–force attacks,” IEEE Trans. Inf. Theory , vol. 66, no. 1, pp. 114–129, 2020

2020
[21]

The likelihood decoder: Error exponents and mismatch,

J. Scarlett, A. Martinez, and A. Guillén i Fàbregas, “The likelihood decoder: Error exponents and mismatch,” in Proc. 2015 IEEE Int. Symp. Inf. Theory (ISIT) , 2015, pp. 86–90

2015
[22]

The generalized stochastic likelihood decoder: Random coding and expurgated bounds,

N. Merhav, “The generalized stochastic likelihood decoder: Random coding and expurgated bounds,” IEEE Trans. Inf. Theory, vol. 63, no. 8, pp. 5039–5051, 2017

2017
[23]

The method of types [information theory],

I. Csiszár, “The method of types [information theory],” IEEE Trans. Inf. Theory , vol. 44, no. 6, pp. 2505–2523, 1998

1998
[24]

Universal noiseless coding,

L. Davisson, “Universal noiseless coding,” IEEE Trans. Inf. Theory , vol. 19, no. 6, pp. 783–795, 1973

1973
[25]

R. B. Ash, Information Theory. New York, NY , USA: Dover, 1990

1990
[26]

A sequential algorithm for the universal coding of finite memory sources,

M. Weinberger, A. Lempel, and J. Ziv, “A sequential algorithm for the universal coding of finite memory sources,” IEEE Trans. Inf. Theory, vol. 38, no. 3, pp. 1002–1014, 1992

1992
[27]

Complexity of strings in the class of Markov sources,

J. Rissanen, “Complexity of strings in the class of Markov sources,” IEEE Trans. Inf. Theory, vol. 32, no. 4, pp. 526–532, 1986

1986
[28]

The error exponent for the noiseless encoding of finite ergodic Markov sources,

L. Davisson, G. Longo, and A. Sgarro, “The error exponent for the noiseless encoding of finite ergodic Markov sources,” IEEE Trans. Inf. Theory , vol. 27, no. 4, pp. 431–438, 1981

1981
[29]

Type classes of context trees,

A. Martín, G. Seroussi, and M. J. Weinberger, “Type classes of context trees,” IEEE Trans. Inf. Theory , vol. 58, no. 7, pp. 4077–4093, 2012

2012
[30]

Gassiat, Universal Coding and Order Identification by Model Selection Methods

E. Gassiat, Universal Coding and Order Identification by Model Selection Methods . Cham, Switzerland: Springer, 2018

2018
[31]

Guessing revisited: A large deviations approach,

M. K. Hanawal and R. Sundaresan, “Guessing revisited: A large deviations approach,” IEEE Trans. Inf. Theory , vol. 57, no. 1, pp. 70–78, 2011

2011
[32]

Enumerative source encoding,

T. Cover, “Enumerative source encoding,” IEEE Trans. Inf. Theory , vol. 19, no. 1, pp. 73–77, 1973

1973
[33]

Randomised attacks on passwords,

M. K. Hanawal and R. Sundaresan, “Randomised attacks on passwords,” Indian Institute of Science, Tech. Rep. TR-PME-2010-11, 2010. [Online]. Available: https://ece.iisc.ac.in/~rajeshs/reprints/TR-PME-2010-11.pdf

2010
[34]

Oblivious distributed guessing,

S. Boztas, “Oblivious distributed guessing,” in 2012 IEEE International Symposium on Information Theory Proceedings , 2012, pp. 2161–2165

2012
[35]

Guessing with limited memory,

W. Huleihel, S. Salamatian, and M. Médard, “Guessing with limited memory,” in 2017 IEEE International Symposium on Information Theory (ISIT) , 2017, pp. 2253–2257

2017
[36]

Mismatched decoding: Error exponents, second-order rates and saddlepoint approximations,

J. Scarlett, A. Martinez, and A. Guillén i Fàbregas, “Mismatched decoding: Error exponents, second-order rates and saddlepoint approximations,” IEEE Trans. Inf. Theory , vol. 60, no. 5, pp. 2647–2666, 2014

2014

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Reliable communication under channel uncertainty,

A. Lapidoth and P. Narayan, “Reliable communication under channel uncertainty,” IEEE Trans. Inf. Theory, vol. 44, no. 6, pp. 2148–2177, 1998

work page 1998

[2] [2]

Universal decoding for channels with memory,

M. Feder and A. Lapidoth, “Universal decoding for channels with memory,” IEEE Trans. Inf. Theory , vol. 44, no. 5, pp. 1726–1745, 1998

work page 1998

[3] [3]

Nonprobabilistic mutual information with memory,

V . D. Goppa, “Nonprobabilistic mutual information with memory,” Problems Control Inf. Theory, vol. 4, pp. 97–102, 1975

work page 1975

[4] [4]

Csiszár and J

I. Csiszár and J. Körner, Information Theory: Coding Theorems for Discrete Memoryless Systems , 2nd ed. Cambridge, UK: Cambridge Univ. Press, 2011

2011

[5] [5]

Universal decoding for finite-state channels,

J. Ziv, “Universal decoding for finite-state channels,” IEEE Trans. Inf. Theory , vol. 31, no. 4, pp. 453–460, 1985

work page 1985

[6] [6]

Universal decoding for memoryless Gaussian channels with a deterministic interference,

N. Merhav, “Universal decoding for memoryless Gaussian channels with a deterministic interference,” IEEE Trans. Inf. Theory, vol. 39, no. 4, pp. 1261–1269, 1993

work page 1993

[7] [7]

On the universality of the LZ-based decoding algorithm,

A. Lapidoth and J. Ziv, “On the universality of the LZ-based decoding algorithm,” IEEE Trans. Inf. Theory, vol. 44, no. 5, pp. 1746–1755, 1998

work page 1998

[8] [8]

Universal composite hypothesis testing: a competitive minimax approach,

M. Feder and N. Merhav, “Universal composite hypothesis testing: a competitive minimax approach,” IEEE Trans. Inf. Theory, vol. 48, no. 6, pp. 1504–1517, 2002

work page 2002