Decoding Short LDPC Codes via BP-RNN Diversity and Reliability-Based Post-Processing

Rosseel, Joachim; Mannoni, Valérian; Fijalkow, Inbar; Savin, Valentin

doi:10.1109/tcomm.2022.3218821

Cited by 8 publications

(5 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, OSD-p post-processing is applied after a decoding diversity of 3 BP-RNNs, denoted D 3 . The construction method of this decoding diversity is detailed in [13]. For each BP-RNN of D 3 , a single neuron is optimized with the parameters described in section IV-A.…”

Section: B Ccsds Codementioning

confidence: 99%

See 1 more Smart Citation

Sets of complementary LLRs to improve OSD post-processing of BP decoding

Rosseel,

Mannoni,

Savin

et al. 2023

2023 12th International Symposium on Topics in Coding (ISTC)

Self Cite

View full text Add to dashboard Cite

This article deals with Ordered Statistics Decoding (OSD) applied to the soft outputs of the Belief Propagation (BP) algorithm. We first model the weighted sum of the a posteriori LLRs across BP decoding iterations into a neuron. The neuron is then trained with the focal loss to compute for each BP decoding failure a set of accumulated Log Likelihood Ratios (LLRs) suited for OSD post-processing. Then, we propose a recursive selection procedure of LLRs sets, for multiple OSD post-processing. This selection is carried out from the sets of a posteriori LLRs calculated at each BP iteration, and from the accumulated LLRs optimized for the OSD, based on their joint probabilities of failure with OSD post-processing. An OSD is then applied to each set of LLRs belonging to the selection. In addition, we propose to reduce the OSD post-processing decoding complexity without significantly degrading its performance. Our results show that this new decoding method provides an effective way to bridge the gap to maximum likelihood decoding for short and long Low Density Parity Check (LDPC) codes.

show abstract

Section: B Ccsds Codementioning

confidence: 99%

“…Employing directly the observed LLRs for the reliability measure has been investigated as well in [12]. The OSD post-processing is also carried out in [13] after a decoding diversity composed of several BP modeled with recurrent neural networks (BP-RNNs), each being trained to decode a different kind of error events.…”

Section: Introductionmentioning

confidence: 99%

Sets of complementary LLRs to improve OSD post-processing of BP decoding

Rosseel,

Mannoni,

Savin

et al. 2023

2023 12th International Symposium on Topics in Coding (ISTC)

Self Cite

View full text Add to dashboard Cite

show abstract

“…40), yet both are surpassed about 0.5dB by the decoder 'NBP-D(10,4,4)' [32], which is slightly more inferior than 'NMS-2-OSD(C-2)-N'. When the order increases to p = 3 for the OSD, the performance of related variants is further improved, among which the 'NMS-2-OSD(D-3)-Y' approaches the SOTA decoder 'D 10 -OSD-2' [33], which is within 0.2dB gap of ML decoding [42],…”

Section: B Decoding Performancementioning

confidence: 99%

“…For (128,64) or (1023,880) code, the DIA procedure accounts for about 220k (31M) Flops with 247 (2071) trainable parameters. In contrast, the SOTA decoder of D 10 -OSD-2(25) [33] requests a bunch of BP-RNN decoders to be trained whose trainable parameters sum up to 10240. Therefore, our approach can greatly relieve the training load.…”

Section: Complexity Analysismentioning

confidence: 99%

“…In [32], at the cost of intensive computation, a neural BP with decimation (NBP-D) was found capable of boosting decoding enormously after ensembling multiple decoding results. Confronted with the fact many BP decoding failures are due to absorbing or trapping sets in code intrinsic structure, a group of RNN models pinpointed for each dominant set was trained individually in [33], and the architecture of interleaving of such a BP-RNN ensemble and OSD made it possible to approach ML performance of short LDPC codes, hence it is regarded as the state of the art (SOTA) in the paper, albeit intensified training load and model complexity.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A bottom‐up design methodology of neural min‐sum decoders for LDPC codes

Luo

et al. 2022

IET Communications

View full text Add to dashboard Cite

It is known the belief propagation variants of linear codes can be readily unrolled as neural networks, after assigning learnable weights on the message-passing edges. Contrary to the conventional top-down training process, where the distillation occurs in the form of pruning or sharing when downsizing model is required, a new bottom-up design methodology to augment performance of the raw min-sum decoder of LDPC codes is proposed, by introducing incrementally a few parameters in the specific positions of corresponding neural network. Then a novel postprocessing method, devised to further improve performance, can cope with decoding failures effectively. In the training process, a simplified scheme of generating training data is presented via exploiting an approximation to the targeted mixture density, and it is found the evaluation of trained parameters converges after sufficient iterations, indicating its generality with an arbitrary designated number of iterations. Lastly, an extensive simulation of three codes carried on the AWGN or Rayleigh fading channels demonstrates the design reaches a good tradeoff of low-complexity and comparable decoding performance.

show abstract