Fig 7 - available via license: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
Content may be subject to copyright.
Source publication
It was known a standard min-sum decoder can be unrolled as a neural network after weighting each edges. We adopt the similar decoding framework to seek a low-complexity and high-performance decoder for a class of finite geometry LDPC codes in short and moderate block lengths. It is elaborated on how to generate high-quality training data effectivel...
Context in source publication
Context 1
... performance, except with a marginal difference at most. More interestingly, the ensemble of SNNMS, UNNMS etc. with T = 5 iterations, all outperform SBP with T = 50 iterations in the waterfall region, even though their performance lags behind the Measured (50) within 0.2dB. The similar conclusions hold for FER comparison as well, as illustrated in Fig. 8. Thus the emergence of neural networks is an opportunity of designing a low-complexity, low-latency, and high-performance decoders. The longer (1023,781) code has a row weight of 32 in its H, and the gap between SBP and Measured (50) is widened to more than 0.3dB at the point of BER=10 −4 as seen in Fig. 9. Despite of that, the ...
