Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

IEEE International Midwest Symposium on Circuits and Systems (MWSCAS)

Federated Transfer Learning for Two-Stage Decoding of LDPC Codes

Linh Nguyen, Quoc Bao Phan, Cheol-Hong Min, Tuy Tan Nguyen

Published Aug 10, 2025• 0 citations
Publisher

Abstract

The performance of fifth-generation (5G) coding techniques, such as low-density parity-check (LDPC) codes, is significantly affected by the error floor, which occurs in high signal-to-noise ratio (SNR) regions. Federated transfer learning (FTL), with its ability to learn from distributed data, can effectively improve error correction in low-SNR environments and extract crucial decoding features for high-SNR conditions. This paper introduces a two-stage decoding method that integrates layered LDPC decoding with FTL to mitigate the error floor and enhance system performance. Experimental results demonstrate that this approach significantly reduces error rates and improves decoding efficiency.

DOI
10.1109/mwscas53549.2025.11244494

Related papers

Federated Transfer Learning for Two-Stage Decoding of LDPC Codes in 6G Networks

Quoc Bao Phan, Linh Nguyen, Tuy Tan Nguyen

The fifth-generation new radio coding techniques, such as low-density parity-check (LDPC) codes, have made significant progress. However, their limitations in the low error floor region prevent them from meeting the stringent bit error rate (BER) targets required for sixth-generation (6G) communications. 6G aims to support services with BER as low as 10 −9 , providing the reliability needed for mission-critical applications. his is particularly crucial for ultra-reliable lowlatency communication. In this paper, we present a two-stage decoding method that combines the strengths of layered LDPC decoding and federated transfer learning (FTL). The layered LDPC decoding stage effectively handles the challenging waterfall region, reducing errors during the initial decoding process. In addition, the FTL model adapts to mitigate the error floor and accelerate its decline. This combination achieves lower error rates and faster performance improvements, reducing the need for retransmissions and enhancing overall system efficiency.

•Nov 12, 2024•0 citations

A Deep Learning-Aided Post-Processing Scheme to Lower the Error Floor of LDPC Codes

Jiajun He

Low-density parity-check (LDPC) codes are widely used in a variety of communication systems because of their low complexity iterative decoding and performance approaching Shannon capacity. However, the existence of error floor seriously hinders the application and development of LDPC codes in related fields. In this paper, a deep learning-aided post-processing scheme (DLAPPS) based on Layered Normalized Min-Sum algorithm (LNMSA) is proposed to effectively lower error floor of LDPC codes. LNMSA is used in the first stage, and in the second stage incorrect frames are re-decoded by processing two types of nodes classified by neural networks. Simulation results show that the proposed DLAPPS can significantly improve error correction performance and lower the error floor of LDPC codes while maintaining a relatively low decoding complexity.

IEEE International Conference on Communication Technology (ICCT)•Oct 28, 2020•4 citations

Boosting Learning for LDPC Codes to Improve the Error-Floor Performance

Hee-Youl Kwak, Dae-Young Yun, Yongjune Kim +2 more

Low-density parity-check (LDPC) codes have been successfully commercialized in communication systems due to their strong error correction capabilities and simple decoding process. However, the error-floor phenomenon of LDPC codes, in which the error rate stops decreasing rapidly at a certain level, presents challenges for achieving extremely low error rates and deploying LDPC codes in scenarios demanding ultra-high reliability. In this work, we propose training methods for neural min-sum (NMS) decoders to eliminate the error-floor effect. First, by leveraging the boosting learning technique of ensemble networks, we divide the decoding network into two neural decoders and train the post decoder to be specialized for uncorrected words that the first decoder fails to correct. Secondly, to address the vanishing gradient issue in training, we introduce a block-wise training schedule that locally trains a block of weights while retraining the preceding block. Lastly, we show that assigning different weights to unsatisfied check nodes effectively lowers the error-floor with a minimal number of weights. By applying these training methods to standard LDPC codes, we achieve the best error-floor performance compared to other decoding methods. The proposed NMS decoder, optimized solely through novel training methods without additional modules, can be integrated into existing LDPC decoders without incurring extra hardware costs. The source code is available at https://github.com/ghy1228/LDPC_Error_Floor .

arXiv:2310.07194•neural information processing systems•Jan 1, 2023•13 citations

Multi-Stage Decoding Scheme with Post-Processing for LDPC Codes to Lower the Error Floors

Beomkyu Shin, Hosung Park, Jong‐Seon No +1 more

In this letter, we propose a multi-stage decoding scheme with post-processing for low-density parity-check (LDPC) codes, which remedies the rapid performance degradation in the high signal-to-noise ratio (SNR) range known as error floor. In the proposed scheme, the unsuccessfully decoded words of the previous decoding stage are re-decoded by manipulating the received log-likelihood ratios (LLRs) of the properly selected variable nodes. Two effective criteria for selecting the probably erroneous variable nodes are also presented. Numerical results show that the proposed scheme can correct most of the unsuccessfully decoded words of the first stage having oscillatory behavior, which are regarded as a main cause of the error floor.

IEICE Transactions on Communications•Jan 1, 2011•6 citations

High Throughput Priority-Based Layered QC-LDPC Decoder with Double Update Queues for Mitigating Pipeline Conflicts

Yunfeng Li, Yingchun Li, Nan Ye +3 more

A high-throughput layered decoder for quasi-cyclic (QC) low-density parity-check (LDPC) codes is required for communication systems. The preferred way to improve the throughput is to insert pipeline stages and increase the operating frequency, which suffers from pipeline conflicts at the same time. A priority-based layered schedule is proposed to keep the updates of log-likelihood ratios (LLRs) as frequent as possible when pipeline conflicts happen. To reduce pipeline conflicts, we also propose double update queues for layered decoders. The proposed double update queues improve the percentage of updated LLRs per iteration. Benefitting from these, the performance loss of the proposed decoder for the fifth generation (5G) new radio (NR) is reduced from 0.6 dB to 0.2 dB using the same quantization compared with the state-of-the-art work. As a result, the throughput of the proposed decoder improved up to 2.85 times when the signal-to-noise ratio (SNR) was equal to 5.9 dB.

Sensors•May 5, 2022•11 citations

CNN-Based Approach for Enhancing 5G LDPC Code Decoding Performance

Sivarama Prasad Tera, Ravikumar Chinthaginjala, Priya Natha +3 more

This paper presents the novel Stochastic decoding-convolutional neural network (SD-CNN) structure, with the goal of enhancing 5G LDPC code’s decoding efficiency in the presence of correlated noise of the channel. Applying the stochastic approach acts as an alternate technique to fixed-point LDPC decoding to enhance the decoder’s hardware efficiency. In order to improve the efficacy of the decoder, we adopted deep learning method such as Convolutional neural network (CNN). CNNs can be employed for denoising purposes in communication systems, where signals may be compromised by diverse forms of noise while being transmitted, with the aim of enhancing signal quality and dependability. The SD-CNN architecture combines a trained CNN with a stochastic decoder of 5G LDPC codes, thereby utilizing the CNN’s capability to accurately estimate channel noise and, consequently, enhance the error correction capabilities of the decoder. The generated output of the trained CNN is then fed back into the stochastic decoder, creating an iterative process between the SD and CNN that ultimately leads to superior decoding performance. For 5G LDPC code word length$N = 3808$, with a base code rate$R = 1/3$, the suggested SD-CNN architecture achieves a BER of$10^{-6}$at 0.6 dB of SNR per bit in the strong correlation of channel noise condition, in comparison to SD, which achieves a BER of$10^{-6}$at 3.5 dB of SNR per bit. The results demonstrate that there is a 2.9 dB improvement.

IEEE Access•Jan 1, 2024•30 citations
The performance of fifth-generation (5G) coding techniques, such as low-density parity-check (LDPC) codes, is significantly affected by the error floor, which occurs in high signal-to-noise ratio (SNR) regions. Federated transfer learning (FTL), with its ability to learn from distributed data, can effectively improve error correction in low-SNR environments and extract crucial decoding features for high-SNR conditions. This paper introduces a two-stage decoding method that integrates layered LDPC decoding with FTL to mitigate the error floor and enhance system performance. Experimental results demonstrate that this approach significantly reduces error rates and improves decoding efficiency.
The fifth-generation new radio coding techniques, such as low-density parity-check (LDPC) codes, have made significant progress. However, their limitations in the low error floor region prevent them from meeting the stringent bit error rate (BER) targets required for sixth-generation (6G) communications. 6G aims to support services with BER as low as 10 −9 , providing the reliability needed for mission-critical applications. his is particularly crucial for ultra-reliable lowlatency communication. In this paper, we present a two-stage decoding method that combines the strengths of layered LDPC decoding and federated transfer learning (FTL). The layered LDPC decoding stage effectively handles the challenging waterfall region, reducing errors during the initial decoding process. In addition, the FTL model adapts to mitigate the error floor and accelerate its decline. This combination achieves lower error rates and faster performance improvements, reducing the need for retransmissions and enhancing overall system efficiency.
Low-density parity-check (LDPC) codes are widely used in a variety of communication systems because of their low complexity iterative decoding and performance approaching Shannon capacity. However, the existence of error floor seriously hinders the application and development of LDPC codes in related fields. In this paper, a deep learning-aided post-processing scheme (DLAPPS) based on Layered Normalized Min-Sum algorithm (LNMSA) is proposed to effectively lower error floor of LDPC codes. LNMSA is used in the first stage, and in the second stage incorrect frames are re-decoded by processing two types of nodes classified by neural networks. Simulation results show that the proposed DLAPPS can significantly improve error correction performance and lower the error floor of LDPC codes while maintaining a relatively low decoding complexity.
Low-density parity-check (LDPC) codes have been successfully commercialized in communication systems due to their strong error correction capabilities and simple decoding process. However, the error-floor phenomenon of LDPC codes, in which the error rate stops decreasing rapidly at a certain level, presents challenges for achieving extremely low error rates and deploying LDPC codes in scenarios demanding ultra-high reliability. In this work, we propose training methods for neural min-sum (NMS) decoders to eliminate the error-floor effect. First, by leveraging the boosting learning technique of ensemble networks, we divide the decoding network into two neural decoders and train the post decoder to be specialized for uncorrected words that the first decoder fails to correct. Secondly, to address the vanishing gradient issue in training, we introduce a block-wise training schedule that locally trains a block of weights while retraining the preceding block. Lastly, we show that assigning different weights to unsatisfied check nodes effectively lowers the error-floor with a minimal number of weights. By applying these training methods to standard LDPC codes, we achieve the best error-floor performance compared to other decoding methods. The proposed NMS decoder, optimized solely through novel training methods without additional modules, can be integrated into existing LDPC decoders without incurring extra hardware costs. The source code is available at https://github.com/ghy1228/LDPC_Error_Floor .
In this letter, we propose a multi-stage decoding scheme with post-processing for low-density parity-check (LDPC) codes, which remedies the rapid performance degradation in the high signal-to-noise ratio (SNR) range known as error floor. In the proposed scheme, the unsuccessfully decoded words of the previous decoding stage are re-decoded by manipulating the received log-likelihood ratios (LLRs) of the properly selected variable nodes. Two effective criteria for selecting the probably erroneous variable nodes are also presented. Numerical results show that the proposed scheme can correct most of the unsuccessfully decoded words of the first stage having oscillatory behavior, which are regarded as a main cause of the error floor.
A high-throughput layered decoder for quasi-cyclic (QC) low-density parity-check (LDPC) codes is required for communication systems. The preferred way to improve the throughput is to insert pipeline stages and increase the operating frequency, which suffers from pipeline conflicts at the same time. A priority-based layered schedule is proposed to keep the updates of log-likelihood ratios (LLRs) as frequent as possible when pipeline conflicts happen. To reduce pipeline conflicts, we also propose double update queues for layered decoders. The proposed double update queues improve the percentage of updated LLRs per iteration. Benefitting from these, the performance loss of the proposed decoder for the fifth generation (5G) new radio (NR) is reduced from 0.6 dB to 0.2 dB using the same quantization compared with the state-of-the-art work. As a result, the throughput of the proposed decoder improved up to 2.85 times when the signal-to-noise ratio (SNR) was equal to 5.9 dB.
This paper presents the novel Stochastic decoding-convolutional neural network (SD-CNN) structure, with the goal of enhancing 5G LDPC code’s decoding efficiency in the presence of correlated noise of the channel. Applying the stochastic approach acts as an alternate technique to fixed-point LDPC decoding to enhance the decoder’s hardware efficiency. In order to improve the efficacy of the decoder, we adopted deep learning method such as Convolutional neural network (CNN). CNNs can be employed for denoising purposes in communication systems, where signals may be compromised by diverse forms of noise while being transmitted, with the aim of enhancing signal quality and dependability. The SD-CNN architecture combines a trained CNN with a stochastic decoder of 5G LDPC codes, thereby utilizing the CNN’s capability to accurately estimate channel noise and, consequently, enhance the error correction capabilities of the decoder. The generated output of the trained CNN is then fed back into the stochastic decoder, creating an iterative process between the SD and CNN that ultimately leads to superior decoding performance. For 5G LDPC code word lengthN=3808N = 3808N=3808, with a base code rateR=1/3R = 1/3R=1/3, the suggested SD-CNN architecture achieves a BER of10−610^{-6}10−6at 0.6 dB of SNR per bit in the strong correlation of channel noise condition, in comparison to SD, which achieves a BER of10−610^{-6}10−6at 3.5 dB of SNR per bit. The results demonstrate that there is a 2.9 dB improvement.