A Model-Driven Deep Learning Method for Normalized Min-Sum LDPC Decoding

来源

Q. Wang, S. Wang, H. Fang, L. Chen, L. Chen and Y. Guo, “A Model-Driven Deep Learning Method for Normalized Min-Sum LDPC Decoding,” 2020 IEEE International Conference on Communications Workshops (ICC Workshops), Dublin, Ireland, 2020, pp. 1-6, doi: 10.1109/ICCWorkshops49005.2020.9145237.

背景

[9] F. Liang, C. Shen, and F. Wu, “An iterative bp-cnn architecture for channel decoding,” IEEE Journal of Selected Topics in Signal Processing,
vol. 12, no. 1, pp. 144–159, 2018.
Author in [9]

use the convolutional neural network (CNN) to estimate the noise
accurately, and then concatenate a conventional LDPC BP decoder to
improve the BER performance.

[10] T. Gruber, S. Cammerer, J. Hoydis, and S. ten Brink, “On deep
learning-based channel decoding,” in 2017 51st Annual Conference on
Information Sciences and Systems (CISS). IEEE, 2017, pp. 1–6.

In [10], a deep-learning-based decoder was proposed for short code to reduce the computational complexity, showing good BER performance
by learning a huge amount of codewords. These BP unrolling based
methods show the promising results for short code decoding [10].

[11] W. Xu, Z. Wu, Y.-L. Ueng, X. You, and C. Zhang, “Improved polar
decoder based on deep learning,” in 2017 IEEE International Workshop
on Signal Processing Systems (SiPS). IEEE, 2017, pp. 1–6.

As the code length grows, training the neural network will require a huge collection of codewords dataset [11].

[12] H. Ye, G. Y. Li, and B.-H. Juang, “Power of deep learning for
channel estimation and signal detection in ofdm systems,” IEEE Wireless
Communications Letters, vol. 7, no. 1, pp. 114–117, 2017.

It is worth mentioning that, the number of information bits k
determines the number of classes (2k) which the network has
to classify [12].

To this end, researchers try to reduce the decoding complexity via reducing the number of iterations or operations.
[13] E. Nachmani, Y. Be’ery, and D. Burshtein, “Learning to decode linear
codes using deep learning,” in 2016 54th Annual Allerton Conference
on Communication, Control, and Computing (Allerton). IEEE, 2016,
pp. 341–346.
[14] E. Nachmani, E. Marciano, D. Burshtein, and Y. Be’ery, “Rnn decoding
of linear block codes,” arXiv preprint arXiv:1702.07560, 2017.

Nachmani et al. assign weights to the edges of the
Tanner graph for Bose-Chaudhuri-Hocquenghem (BCH) codes
[13] [14], which achieves similar decoding performance with
fewer iterations compared with conventional BP decoders.

[15] L. Lugosch and W. J. Gross, “Neural offset min-sum decoding,” in 2017
IEEE International Symposium on Information Theory (ISIT). IEEE,
2017, pp. 1361–1365.

However, a large number of weights in these methods result in a high decoding complexity, making it impossible in long code applications. Considering the many logarithmic and
multiplicative computations in BP decoding, Lugosch et al.
propose a neural offset min-sum (NOMS) decoding algorithm
without any multiplication operations [15], which performs
well on hardware.

But existing methods designed for regular
codes can hardly be applied to long code, which refers to codes
with length longer than one hundred comparing with short
codes with length of several tens
In a word, the complexity of computation hinders the application of neural networks for long code decoding.

作者主要贡献

1. 提出NNMS(neural normalized min-sum)解码算法,为实现长码LDPC 的高准确率和低复杂度的译码
2. 算法主要想法是将传统的BP(信念传播)算法中,校验节点和变量节点之间的迭代译码的过程展开为前馈神经网络.
3. 另外由于NNMS 需要大量的乘法操作, 提出了shared neural normalized minsum (SNNMS) decoding network,通过在每一层使用相同的correction factors 以降低可学习的correction factors 的数目.

[16] J. Liu, S. Jing, X. You, and C. Zhang, “A merged bp decoding
algorithm for polar-ldpc concatenated codes,” in 2017 22nd International
Conference on Digital Signal Processing (DSP). IEEE, 2017, pp. 1–5.
[17] N. Yang, S. Jing, A. Yu, X. Liang, Z. Zhang, X. You, and C. Zhang,
“Reconfigurable decoder for ldpc and polar codes,” in 2018 IEEE
International Symposium on Circuits and Systems (ISCAS). IEEE, 2018,
pp. 1–5.
[18] H. He, S. Jin, C.-K. Wen, F. Gao, G. Y. Li, and Z. Xu, “Modeldriven deep learning for physical layer communications,” IEEE Wireless
Communications, 2019.