Pith. sign in

REVIEW 3 major objections 5 minor 45 references

GNN-based Auto-Encoder for Short Linear Block Codes: A DRL Approach

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Joint machine-learning design of short codes and their decoders beats LDPC and BCH at short block lengths, the paper claims.

desk verdict Serious joint code/decoder design with honest failure reporting, but the headline MLD gains rest on an unverified reward-proxy assumption. read the letter →

arxiv 2412.02053 v1 pith:7U3FPPCI submitted 2024-12-03 cs.LG cs.ITmath.IT

Kou Tian , Chentao Yue , Changyang She , Yonghui Li , Branka Vucetic This is my paper · ORCID
classification cs.LGcs.ITmath.IT MSC 94B0594B3568T07
keywords channelcodingshortlinearblockcodesdeepreinforcementlearninggraphneuralnetworksbeliefpropagationauto-encoderURLLCTanner
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 claims that a short linear block code and its decoder can be designed together by machine learning, and that the resulting pair beats conventional coding at the block lengths used in ultra-reliable low-latency communications. The authors build a reinforcement-learning agent that generates parity-check matrices by flipping entries of an initial matrix, and pair it with a graph-neural-network decoder whose per-edge weights suppress unreliable messages from short cycles. Trained jointly in alternation, the code and decoder reach coding gains of about 1 dB over a standard (32,16) LDPC code decoded with belief propagation, 0.63 dB over the same LDPC code decoded with maximum-likelihood decoding at 30 iterations, and 1.81 dB over a (63,45) BCH code with BP. If true, this would mean learned codes can replace hand-designed algebraic codes in short-block regimes while keeping decoding complexity similar to BP.

What carries the argument

The argument runs on three coupled objects. First, the DRL code designer treats the parity-check matrix as a lattice graph and each entry as a node; a message-passing matrix neural network (the actor and critic) outputs a flip probability for every entry, and the agent explores the space of full-rank matrices. Its reward is either the simulated BER at 6 dB with an 8-iteration BP decoder, $r_t^d = |\ln \epsilon_b(S_{t+1})|$, or a structure reward $r_t^s = d_{\min}/\alpha_d + \alpha_c/(c_4 + \alpha_c)$ that rewards large minimum Hamming distance and few 4-cycles, with $\alpha_d = 8$ and $\alpha_c = 500$. Second, the EW-GNN decoder runs BP-style message passing on the Tanner graph but multiplies each check-to-variable message by a learned edge weight $w^{(t)}_{u_j \to v_i}=g(|\mu^{(t)}_{u_j\to v_i}|, \delta^{(t)}(\mu_{u_j\to v_i}), \delta^{(t-1)}(\mu_{v_i\to u_j}), \delta^{(t-1)}(h_{v_i})|\theta_g)$, with the $\tanh^{-1}$ in the check update clipped for differentiability; because all edges share the same MLP, the parameter count is independent of code length. Third, alternating training lets the decoder's BER become the reward for the next code-search iteration, so code and decoder co-adapt, and the paper reports convergence in about three iterations.

What would settle it

Compute the minimum Hamming distance of the auto-encoder code $H^*_{AE}$ and evaluate its BER at SNRs well below and above 6 dB (e.g., 3 dB and 8 dB) with the EW-GNN decoder and with MLD; if the code's low-density, low-cycle structure makes it worse than the reference LDPC under MLD away from the training point—as already happens for the BER-reward-only code $H^*_d$—then the single-SNR reward is not a valid proxy and the headline gains are operating-point artifacts.

Watch

Extended reading notes

Core claim

The central claim is that the proposed auto-encoder—a DRL-based code designer that searches parity-check matrices plus an edge-weighted GNN decoder—produces codes whose error-correction performance surpasses several classical schemes at short block lengths. In the (32,16) case, the learned code with the EW-GNN decoder gains about 1 dB over a standard LDPC code with BP, and gains 0.63 dB over LDPC with maximum-likelihood decoding when the decoder is run for 30 iterations; the learned code also outperforms LDPC under MLD by 1.28 dB. For (63,45), the auto-encoder gains 1.81 dB over BCH with BP. The paper further claims that the EW-GNN decoder, once trained on one linear block code, can decode other codes of different lengths and rates without retraining, and that the iterative joint training converges within about three iterations.

Load-bearing premise

The load-bearing premise is that the reward used during code search—simulated BER at one SNR (6 dB) with a fixed 8-iteration BP decoder, or a hand-weighted combination of minimum distance and 4-cycle count—correctly ranks code quality at all operating SNRs and under the final decoder.

Editorial extensions

If this is right

  • Short-block URLLC links could use learned codes that outperform classical LDPC and BCH codes under practical iterative decoding, narrowing the gap to finite-length bounds at these block lengths.
  • A single trained decoder can be reused for different code lengths and rates, so a system could switch codes dynamically without retraining the receiver.
  • The learned code plus EW-GNN can beat maximum-likelihood decoding of a standard LDPC code at high SNR while keeping BP-like per-iteration complexity, relaxing the usual performance-complexity trade-off.
  • Joint encoder-decoder co-design converges in a few alternations, suggesting that end-to-end optimization is tractable for practical block lengths rather than only toy sizes.

Reading between the lines

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

  • The reward is a single-SNR, single-decoder proxy: the paper's own results show that the BER-reward-only code has minimum Hamming distance 3 and is worse than LDPC under MLD. An analogous failure for the auto-encoder code would mean the reported gains are tied to the 6 dB operating point and the 8-iteration BP/EW-GNN decoder, not to intrinsic code quality.
  • The structure-reward constants ($\alpha_d=8$, $\alpha_c=500$, and cycle counting limited to 4-cycles) are hand-picked; tuning or replacing them with differentiable estimates of minimum distance and girth could change the discovered codes, and the paper does not study this sensitivity.
  • Because EW-GNN transfers across code lengths, the same machinery could be probed on much longer codes or on nonbinary or quantum LDPC families, but the paper only demonstrates transfer up to length 256 at rate 1/2.
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

3 major / 5 minor

Summary. The paper proposes an end-to-end channel encoding/decoding system, termed a GNN-based auto-encoder, for short linear block codes. A DRL-based agent, trained with DDPG, generates parity-check matrices by flipping entries of an initial matrix; the agent can be rewarded either by a simulated BER at a single SNR (Eq. 7) or by a structural reward involving minimum distance and short cycles (Eq. 8). An edge-weighted GNN decoder (EW-GNN) is introduced, which aligns with belief propagation but learns multiplicative edge weights to suppress unreliable messages. The designer and decoder are trained iteratively, alternating between code design and decoder fine-tuning. Experiments report BER gains over CCSDS LDPC and BCH codes, as well as over the NBP-AE baseline, for (32,16) and (63,45) codes, and demonstrate decoder transferability across code lengths and rates without retraining.

Significance. If the reported gains are reproducible, the work advances learning-based code design by combining a scalable GNN decoder with a DRL-based parity-check matrix search, and it explicitly targets the short-block-length regime where BP and LDPC codes are known to be suboptimal. The paper has notable strengths: the evaluation protocol is clearly described (10,000 bit errors per point), the decoder's transferability across code lengths is directly demonstrated (Figs. 8-9), and the authors honestly disclose negative results, including that the BP-BER-rewarded code H*_d is MLD-inferior to LDPC (Table II) and that the learned (63,45) code is MLD-inferior to BCH (Fig. 11). These admissions increase confidence in the empirical protocol. However, the central claim that the auto-encoder significantly surpasses LDPC with MLD depends on the MLD curves for the learned code, whose intrinsic code quality (minimum distance, weight enumerator) is never reported and whose MLD implementation is not described.

major comments (3)
  1. [VI-A and Table II] The paper's own Table II shows that the BP-BER reward of Eq. (7), evaluated at a single SNR (6 dB) with 8-iteration BP, selects a code H*_d whose minimum distance is 3 and whose MLD BER at 5 dB (2.64e-4) is worse than the CCSDS LDPC baseline (2.29e-4). The auto-encoder code H*_AE, whose headline MLD gains appear in Figs. 10 and 12, is selected by the same type of decoder-BER reward, with EW-GNN replacing BP in the iterative loop of Section V. The paper reports only the number of 4-cycles for H*_AE (77 in Section VI-C1) and never reports dmin, the weight enumerator, or an independently verified MLD curve. Without this information, the claimed 1.28 dB MLD gain over LDPC (Fig. 10) and 0.63 dB gain over LDPC with MLD (Fig. 10) could be operating-point artifacts of matching EW-GNN's fixed point at the reward SNR rather than intrinsic code quality. Please add dmin(H*_AE) and a validation of the MLD results against an independent MLD implementation or known code database, such as [39].
  2. [VI-C1 and Fig. 10] The MLD curves for H*_AE are obtained without any description of the maximum-likelihood decoding algorithm used. The paper states 'We also test the MLD performance' but does not specify whether this is brute-force nearest-codeword search, trellis-based decoding, or some other exact MLD method, nor does it describe how decoding errors are counted (message bits or codeword bits). This matters because H*_AE is a non-standard code and its MLD curve cannot be cross-checked by the reader without either the parity-check matrix or the decoding algorithm. Please specify the MLD implementation, its complexity, and provide a sanity check by reproducing a known MLD curve (e.g., for the CCSDS LDPC baseline) with the same implementation.
  3. [VI-C3 and Fig. 13] Fig. 13 is used to claim that the coding gain between H*_2 and H*_3 'diminishes' and to attribute this to EW-GNN scalability. However, H*_1 is designed with a BP-based reward, whereas H*_2 and H*_3 are designed with the EW-GNN-based reward; the comparison therefore conflates a change of reward function with the convergence of the iterative training. A more direct convergence test would be to run more than three iterations or to report the reward values and code properties (in particular dmin and the cycle distribution) at each iteration. Please also clarify whether the 'MLD' curves in Fig. 13 are computed with the same decoding algorithm as in Fig. 10 and whether the same training setup was used for all three iterations.
minor comments (5)
  1. [Section III-A] The definition 'SNR = 1/σ_n^2' is incomplete for a BPSK AWGN channel; please state whether this is Es/N0 or Eb/N0 and clarify the normalization of the symbol energy, since the coding gain comparisons in Section VI depend on the SNR convention.
  2. [Figure 3 and Table I] Figure 3 states that the flipping threshold α_f is 0.5, while Table I reports α_f = 0.3; please align the illustrative example with the actual hyper-parameter setting.
  3. [Section VI-C1] The text says H*_AE has '77 4-step cycles' and that NBP-AE has '172 short cycles'; the metric for NBP-AE is not defined in Table II or elsewhere. Please report the same cycle metric for all compared codes for a fair comparison.
  4. [Eq. (8)] The structural reward r_t^s is not normalized to a scale comparable with the BER reward r_t^d in Eq. (7); please clarify whether the two rewards are ever mixed in one training run and how their relative scales are handled in the DDPG training.
  5. [Abstract and Fig. 7 caption] There are minor language issues: 'auto-encoder based' should be hyphenated, and the caption of Fig. 7 contains 'Peformance' instead of 'Performance'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the code-design results are empirical optimization outcomes with independent MLD and cross-code evaluations, not predictions forced by construction.

full rationale

The paper's only load-bearing derivation is the DRL search over parity-check matrices, with rewards r_t^d = |ln(epsilon_b(S_{t+1}))| (Eq. 7) and r_t^s (Eq. 8). Reporting the resulting codes' BER is reporting the objective of the search, but the paper does not present this as an independent prediction: it is a simulation result of a designed system. The key headline comparisons are not forced by the reward definition. (i) H*_s is selected with the structure reward, yet its MLD gain of 0.83 dB over LDPC (Section VI-A, Fig. 7) is an independent test; (ii) H*_AE is selected with EW-GNN BER rewards, yet its MLD gains (1.28 dB over LDPC, 0.93 dB over NBP-AE, Fig. 10) and its (63,45) MLD behavior (Fig. 11) are not part of the optimized objective; (iii) the EW-GNN decoder's cross-length and cross-rate generalization (Figs. 8-9) is verified against external BP/NBP baselines. The paper explicitly acknowledges the reward-proxy limitation: the BP-BER reward produces H*_d with MHD=3 and MLD-inferior BER to LDPC (Table II), so the optimization objective is not identical to the reported evaluation. The only self-citation, [20], supports a complexity/scalability claim that is also demonstrated by simulations in the present paper; no load-bearing result rests on a self-citation chain, and no uniqueness theorem or ansatz is imported from the authors' prior work. Hence the derivation chain is self-contained and non-circular.

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

The central claims rest on the AWGN channel model, the single-SNR BER reward and hand-chosen structure reward constants that define what the DRL agent optimizes, the heuristic algorithmic-alignment assumption connecting GNN training to BP behavior, the external cycle-counting method, and an unspecified MLD procedure for the designed (63,45) code. The fitted quantities are the neural network weights and the selected parity-check matrices themselves, which are the output of the training procedure rather than independent measurements, so the ledger contains no parameters fit against external benchmark data.

free parameters (6)
  • flipping threshold alpha_f = 0.3
    Element-flip threshold in the MDP state transition (Eq. 6 and Table I); chosen by hand, it controls how far the designed code can move from the seed matrix within the 25 steps per episode.
  • structure reward constants alpha_d, alpha_c = alpha_d=8, alpha_c=500
    Balance minimum Hamming distance against 4-cycle count in Eq. (8); chosen by hand with no sensitivity analysis, and the paper does not report how results vary with them.
  • BER reward operating SNR = 6 dB
    The reward in Eq. (7) is measured by BP simulation at a single SNR; the claim that designed codes perform well across a wide SNR range rests on this one operating point (Table I).
  • EW-GNN clip factor alpha = 1e-32 for BCH, 1e-7 for LDPC
    Replaces the unbounded tanh^-1 in the BP update (Eq. 21) to make training differentiable; tuned separately per code family (Table III) and affects the numerical behavior of message passing.
  • cycle length lambda in structure reward = 4
    Only 4-cycles are penalized in Eq. (8); 6-cycles and longer are ignored by construction even though Table II reports 6-cycle counts.
  • number of training iterations T = 8
    EW-GNN is trained with T=8 iterations; gains reported at T=30 are extrapolations of the T=8-trained model, and the clip factor ensures convergence at larger T.
assumptions (6)
  • domain assumption AWGN channel with known noise power and BPSK mapping; LLR computed as in Eq. (2)
    All reward measurements and decoder training assume this channel model and exact knowledge of sigma_n^2 (Sections II-A3 and III-A); if the noise variance is mismatched, the LLR inputs to EW-GNN and the BER reward are biased.
  • ad hoc to paper Single-SNR BER reward generalizes to code quality across all SNRs
    The DRL reward (Eq. 7) is simulated BER at 6 dB with 8-iteration BP only; Table II shows this proxy yields H*_d with MHD=3, MLD-inferior to LDPC, so the premise is contradicted in the asymptotic regime for one reward branch.
  • domain assumption Algorithmic alignment of GNN and BP on Tanner graphs justifies EW-GNN training behavior
    Invoked in Section IV via [27] as the design justification; it is a heuristic argument that a GNN with the same permutation invariance can learn BP-like message passing, not a theorem guaranteeing the reported gains or the transfer to longer codes.
  • standard math Full-rank parity-check matrices define valid linear codes; rank is enforced by the reward
    Algorithm 1 returns zero reward for rank-deficient states, and systematic-form Gaussian elimination (Section II-A1) is standard linear algebra over F2.
  • domain assumption Cycle counts computed by the algorithm in [40] are exact and comparable across codes
    Table II and the comparison of 77 versus 172 four-cycles in Section VI-C1 rely on this external counting method; errors in cycle counts would affect the structure reward and the paper's explanation of the gains.
  • ad hoc to paper The plotted MLD results for the designed (63,45) code are exact
    Fig. 11 shows an MLD curve for a k=45 learned code without stating the algorithm; exhaustive MLD over 2^45 codewords is infeasible, so the claim depends on an unspecified exact or near-exact decoding procedure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GNN-based Auto-Encoder for Short Linear Block Codes: A DRL Approach." pith.science (2026). https://pith.science/paper/7U3FPPCI

@misc{pith2026241202053,
  author       = {Pith},
  title        = {Pith review of: GNN-based Auto-Encoder for Short Linear Block Codes: A DRL Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7U3FPPCI}},
  note         = {Machine review of arXiv:2412.02053}
}
read the original abstract

This paper presents a novel auto-encoder based end-to-end channel encoding and decoding. It integrates deep reinforcement learning (DRL) and graph neural networks (GNN) in code design by modeling the generation of code parity-check matrices as a Markov Decision Process (MDP), to optimize key coding performance metrics such as error-rates and code algebraic properties. An edge-weighted GNN (EW-GNN) decoder is proposed, which operates on the Tanner graph with an iterative message-passing structure. Once trained on a single linear block code, the EW-GNN decoder can be directly used to decode other linear block codes of different code lengths and code rates. An iterative joint training of the DRL-based code designer and the EW-GNN decoder is performed to optimize the end-end encoding and decoding process. Simulation results show the proposed auto-encoder significantly surpasses several traditional coding schemes at short block lengths, including low-density parity-check (LDPC) codes with the belief propagation (BP) decoding and the maximum-likelihood decoding (MLD), and BCH with BP decoding, offering superior error-correction capabilities while maintaining low decoding complexity.

Figures

Figures reproduced from arXiv: 2412.02053 by the authors.

Figure 1
Figure 1. A parity-check matrix in systematic form and its corresponding Tanner [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. System model. neural networks to simultaneously learn a deterministic policy and an action-value function. The actor network determines the optimal actions, and the critic network estimates the Q-value, i.e., the long-term reward. DDPG has more stable and faster convergence than traditional RL by combining policy gradient for optimizing policies and Q-learning for value estimation. 2) Graph Neural Network: GNN is de… view at source ↗
Figure 4
Figure 4. The block diagram of the DRL-based code designer. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (9 more)
Figure 5
Figure 5. Figure 5: Illustration of updating the node embedding in the GNN [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The iterative training phase of the proposed GNN-DRL auto-encoder [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: BER performance for designed (32, 16) linear block codes. 10,000 bit errors are collected. For different coding schemes, we define coding gain as the difference in SNRs required to achieve the same target BER at high SNRs. A. DRL-based Neural Code Designer We evaluate …
Figure 8
Figure 8. Figure 8: BER performance for BCH codes of length n = 63. EW-GNN is only trained with the (63, 51) BCH code and T = 8. 1 2 3 4 5 6 7 8 SNR(dB) 10-6 10-4 10-2 BER (32,16) LDPC BP (T=8) NBP (T=8) EW-GNN (T=8) BP (T=30) NBP (T=30) EW-GNN (T=30) 1 2 3 4 5 SNR(dB) 10-6 10-4 10-2 BER …
Figure 9
Figure 9. Figure 9: BER performance for LDPC codes at rate 1/2. EW-GNN is only trained with the (32, 16) LDPC code and T = 8. 1 2 3 4 5 6 7 8 SNR(dB) 10-8 10-6 10-4 10-2 BER Performance of (32,16) Codes LDPC code BP(T=8) MLD NBP-AE code GNBP(T=8) MLD H *AE code EW-GNN(T=8) EW-GNN(T=30) ML…
Figure 10
Figure 10. Figure 10: BER performance for designed (32, 16) linear block codes. H∗ AE code is generated by our auto-encoder scheme with three training iterations. coding scenarios, making them suitable for a wide range of applications in wireless communications. C. GNN-DRL Auto-encoder We …
Figure 11
Figure 11. Figure 11: BER performance for designed (63, 45) linear block codes. H∗ AE code is generated by our auto-encoder scheme with three training iterations. 1 2 3 4 5 6 7 8 SNR(dB) 10-8 10-6 10-4 10-2 BER Performance of (32,16) Codes LDPC code BP(T=8) EW-GNN(T=8) MLD H * d code BP(T=…
Figure 12
Figure 12. Figure 12: Evaluate the relative contributions of the DRL-based code designer [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 6
Figure 6. Figure 6: As shown in Fig. 13, [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 38 canonical work pages

  1. [39]

    Database of Channel Codes and ML Simulation Results,

    M. Helmling, S. Scholl, F. Gensheimer, T. Dietz, K. Kraft, S. Ruzika, and N. Wehn, “Database of Channel Codes and ML Simulation Results,” www.uni-kl.de/channel-codes, 2019

    2019

  2. [1]

    Short block-length codes for ultra-reliable low latency communications,

    M. Shirvanimoghaddam, M. S. Mohammadi, R. Abbas, A. Minja, C. Yue, B. Matuz, G. Han, Z. Lin, W. Liu, Y . Li, S. Johnson, and B. Vucetic, “Short block-length codes for ultra-reliable low latency communications,” IEEE Commun. Mag. , vol. 57, no. 2, pp. 130–137, Feb. 2019

    work page 2019

  3. [2]

    Efficient decoders for short block length codes in 6g urllc,

    C. Yue, V . Miloslavskaya, M. Shirvanimoghaddam, B. Vucetic, and Y . Li, “Efficient decoders for short block length codes in 6g urllc,” IEEE Commun. Mag. , vol. 61, no. 4, pp. 84–90, Apr. 2023

    work page 2023

  4. [3]

    NR; Multiplexing and channel coding (Release 16), v16.0.0,

    “NR; Multiplexing and channel coding (Release 16), v16.0.0,” 3rd Generation Partnership Project (3GPP), Tech. Rep. TS 38.212, Jan. 2020

    work page 2020

  5. [4]

    Channel coding rate in the finite blocklength regime,

    Y . Polyanskiy, H. V . Poor, and S. Verdu, “Channel coding rate in the finite blocklength regime,” IEEE Trans. Inf. Theory , vol. 56, no. 5, pp. 2307–2359, May 2010

    2010

  6. [5]

    Deep learning for intelligent wireless net- works: A comprehensive survey,

    Q. Mao, F. Hu, and Q. Hao, “Deep learning for intelligent wireless net- works: A comprehensive survey,” IEEE Commun. Surv. Tutor ., vol. 20, no. 4, pp. 2595–2621, 4th Quart. 2018

    work page 2018

  7. [6]

    Deep learning for wireless communications: An emerging interdisciplinary paradigm,

    L. Dai, R. Jiao, F. Adachi, H. V . Poor, and L. Hanzo, “Deep learning for wireless communications: An emerging interdisciplinary paradigm,” IEEE Wireless Commun. , vol. 27, no. 4, pp. 133–139, Aug. 2020

    work page 2020

  8. [7]

    Deep mimo detection,

    N. Samuel, T. Diskin, and A. Wiesel, “Deep mimo detection,” in Proc. IEEE SPA WC, 2017, pp. 1–5

    work page 2017

Show all 45 references
  1. [8]

    Learning to optimize: Training deep neural networks for wireless resource management,

    H. Sun, X. Chen, Q. Shi, M. Hong, X. Fu, and N. D. Sidiropoulos, “Learning to optimize: Training deep neural networks for wireless resource management,” in Proc. IEEE SPA WC, 2017, pp. 1–6

    2017

  2. [9]

    Learn codes: Inventing low-latency codes via recurrent neural networks,

    Y . Jiang, H. Kim, H. Asnani, S. Kannan, S. Oh, and P. Viswanath, “Learn codes: Inventing low-latency codes via recurrent neural networks,” in Proc. IEEE ICC , 2019, pp. 1–7

    2019

  3. [10]

    Deepcode: Feedback codes via deep learning,

    H. Kim, Y . Jiang, S. Kannan, S. Oh, and P. Viswanath, “Deepcode: Feedback codes via deep learning,”IEEE J. Sel. Areas Inf. Theory, vol. 1, no. 1, pp. 194–206, May 2020

    2020

  4. [11]

    Deep learning- based polar code design,

    M. Ebada, S. Cammerer, A. Elkelesh, and S. ten Brink, “Deep learning- based polar code design,” in Proc. IEEE Annu. Allerton Conf. Commun. Control Comput., 2019, pp. 177–183

    2019

  5. [12]

    Ai coding: Learning to construct error correction codes,

    L. Huang, H. Zhang, R. Li, Y . Ge, and J. Wang, “Ai coding: Learning to construct error correction codes,” IEEE Trans. Commun. , vol. 68, no. 1, pp. 26–39, Jan. 2020

    2020

  6. [13]

    On deep learning- based channel decoding,

    T. Gruber, S. Cammerer, J. Hoydis, and S. t. Brink, “On deep learning- based channel decoding,” in Proc. IEEE CISS , 2017, pp. 1–6

    2017

  7. [14]

    Learning to decode linear codes using deep learning,

    E. Nachmani, Y . Be’ery, and D. Burshtein, “Learning to decode linear codes using deep learning,” in Proc. IEEE Annu. Allerton Conf. Com- mun. Control Comput. , 2016, pp. 341–346

    2016

  8. [15]

    Deep learning methods for improved decoding of linear codes,

    E. Nachmani, E. Marciano, L. Lugosch, W. J. Gross, D. Burshtein, and Y . Be’ery, “Deep learning methods for improved decoding of linear codes,” IEEE J. Sel. Topics Signal Process. , vol. 12, no. 1, pp. 119–131, Feb. 2018

    2018

  9. [16]

    Hyper-graph-network decoders for block codes,

    E. Nachmani and L. Wolf, “Hyper-graph-network decoders for block codes,” in Proc. NeurIPS, 2019, pp. 2326–2336

    2019

  10. [17]

    Pruning and quantizing neural belief propagation decoders,

    A. Buchberger, C. H ¨ager, H. D. Pfister, L. Schmalen, and A. Graell i Amat, “Pruning and quantizing neural belief propagation decoders,” IEEE J. Sel. Areas Commun. , vol. 39, no. 7, pp. 1957–1966, Jul. 2021

    1957

  11. [18]

    Neural decoding with optimization of node activations,

    E. Nachmani and Y . Be’ery, “Neural decoding with optimization of node activations,” IEEE Commun. Lett. , vol. 26, no. 11, pp. 2527–2531, Nov. 2022

    2022

  12. [19]

    Graph neural networks for channel decoding,

    S. Cammerer, J. Hoydis, F. A. Aoudia, and A. Keller, “Graph neural networks for channel decoding,” in Proc. IEEE GC Wkshps , 2022, pp. 486–491

    2022

  13. [20]

    A scalable graph neural network decoder for short block codes,

    K. Tian, C. Yue, C. She, Y . Li, and B. Vucetic, “A scalable graph neural network decoder for short block codes,” in Proc. IEEE ICC , 2023, pp. 1268–1273

    2023

  14. [21]

    Neural belief propagation auto-encoder for linear block code design,

    G. Larue, L.-A. Dufrene, Q. Lampin, H. Ghauch, and G. R.-B. Othman, “Neural belief propagation auto-encoder for linear block code design,” IEEE Trans. Commun. , vol. 70, no. 11, pp. 7250–7264, Nov. 2022

    2022

  15. [22]

    An introduction to deep learning for the physical layer,

    T. O’Shea and J. Hoydis, “An introduction to deep learning for the physical layer,” IEEE Trans. Cogn. Commun. Netw. , vol. 3, no. 4, pp. 563–575, Dec. 2017

    2017

  16. [23]

    Circular convolutional auto-encoder for channel coding,

    H. Ye, L. Liang, and G. Y . Li, “Circular convolutional auto-encoder for channel coding,” in Proc. IEEE SPA WC, 2019, pp. 1–5

    2019

  17. [24]

    Turbo autoencoder: Deep learning based channel codes for point-to- point communication channels,

    Y . Jiang, H. Kim, H. Asnani, S. Kannan, S. Oh, and P. Viswanath, “Turbo autoencoder: Deep learning based channel codes for point-to- point communication channels,” in Proc. NeurIPS, vol. 32, 2019

    2019

  18. [25]

    Autoencoder-based error correction coding for one-bit quantization,

    E. Balevi and J. G. Andrews, “Autoencoder-based error correction coding for one-bit quantization,” IEEE Trans. Commun , vol. 68, no. 6, pp. 3440–3451, Jun. 2020

    2020

  19. [26]

    High-dimensional data analysis: The curses and blessings of dimensionality,

    D. L. Donoho et al. , “High-dimensional data analysis: The curses and blessings of dimensionality,” AMS Math. Chall. Lect. , vol. 1, no. 2000, p. 32, 2000

    2000

  20. [27]

    What can neural networks reason about?

    K. Xu, J. Li, M. Zhang, S. S. Du, K. ichi Kawarabayashi, and S. Jegelka, “What can neural networks reason about?” in Proc. ICLR, 2020

    2020

  21. [28]

    A heuristic search for good low-density parity-check codes at short block lengths,

    Y . Mao and A. Banihashemi, “A heuristic search for good low-density parity-check codes at short block lengths,” in Proc. IEEE ICC , vol. 1, 2001, pp. 41–44

    2001

  22. [29]

    Iterative decoding of binary block and convolutional codes,

    J. Hagenauer, E. Offer, and L. Papke, “Iterative decoding of binary block and convolutional codes,” IEEE Trans. Inf. Theory , vol. 42, no. 2, pp. 429–445, Mar. 1996

    1996

  23. [30]

    Factor graphs and the sum-product algorithm,

    F. Kschischang, B. Frey, and H.-A. Loeliger, “Factor graphs and the sum-product algorithm,” IEEE Trans. Inf. Theory , vol. 47, no. 2, pp. 498–519, Feb. 2001

    2001

  24. [31]

    Multiple-bases belief-propagation for decoding of short block codes,

    T. Hehn, J. B. Huber, S. Laendner, and O. Milenkovic, “Multiple-bases belief-propagation for decoding of short block codes,” in Proc. IEEE ISIT, 2007, pp. 311–315

    2007

  25. [32]

    Continuous control with deep reinforcement learning,

    T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y . Tassa, D. Silver, and D. Wierstra, “Continuous control with deep reinforcement learning,” in Proc. ICLR, 2016

    2016

  26. [33]

    Deterministic policy gradient algorithms,

    D. Silver, G. Lever, N. Heess, T. Degris, D. Wierstra, and M. Riedmiller, “Deterministic policy gradient algorithms,” in Proc. ICML , vol. 32. PMLR, 2014, pp. 387–395

    2014

  27. [34]

    Human-level control through deep reinforcement learn- ing,

    V . Mnih et al., “Human-level control through deep reinforcement learn- ing,” Nature, vol. 518, no. 7540, pp. 529–533, Feb. 2015

    2015

  28. [35]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A comprehensive survey on graph neural networks,” IEEE Trans. Neural Netw. Learn. Syst. , vol. 32, no. 1, pp. 4–24, Jan. 2021

    2021

  29. [36]

    Neural message passing for quantum chemistry,

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, and G. E. Dahl, “Neural message passing for quantum chemistry,” in Proc. ICML . PMLR, 2017, p. 1263–1272

    2017

  30. [37]

    Graph neural networks for wireless communications: From theory to practice,

    Y . Shen, J. Zhang, S. H. Song, and K. B. Letaief, “Graph neural networks for wireless communications: From theory to practice,” IEEE Trans. Wireless Commun., vol. 22, no. 5, pp. 3554–3569, May 2023

    2023

  31. [38]

    Bounds on the minimum distance of linear codes and quantum codes,

    M. Grassl, “Bounds on the minimum distance of linear codes and quantum codes,” 2007

    2007

  32. [40]

    Message-passing algorithms for counting short cycles in a graph,

    M. Karimi and A. H. Banihashemi, “Message-passing algorithms for counting short cycles in a graph,” IEEE Trans. Commun., vol. 61, no. 2, pp. 485–495, Feb. 2013

    2013

  33. [41]

    R. S. Sutton and A. G. Barto, Reinforcement learning: An introduction . MIT press, 2018

    2018

  34. [42]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in Proc. ICLR, 2015

    2015

  35. [43]

    Improving belief propa- gation on graphs with cycles,

    M. Yazdani, S. Hemati, and A. Banihashemi, “Improving belief propa- gation on graphs with cycles,” IEEE Commun. Lett. , vol. 8, no. 1, pp. 57–59, Jan. 2004

    2004

  36. [44]

    Informed decoding algorithms of ldpc codes based on dynamic selection strategy,

    X. Liu, Z. Zhou, R. Cui, and E. Liu, “Informed decoding algorithms of ldpc codes based on dynamic selection strategy,” IEEE Trans. Commun., vol. 64, no. 4, pp. 1357–1366, Apr. 2016

    2016

  37. [45]

    Short block length LDPC codes for TC synchronization and channel coding,

    “Short block length LDPC codes for TC synchronization and channel coding,” Consultative Committee for Space Data Systems (CCSDS), Tech. Rep. 231.1-O-1, Apr. 2015

    2015

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.