A bottom‐up design methodology of neural min‐sum decoders for LDPC codes. Issue 3 (14th December 2022)

• Record Type: Journal Article
• Authors: Li, Guangwen; Yu, Xiao; Luo, Yuan; Wei, Guangfen
• Is Part Of: IET communications. Volume 17:Issue 3(2023)
• Page Start: 377
• Page End: 386
• Publication Date: 2022-12-14
• Subjects: belief propagation -- deep learning -- min‐sum -- neural network -- training; Telecommunication systems -- Periodicals; Speech processing systems -- Periodicals; 621.38205
• DOI: 10.1049/cmu2.12547
• Languages: English
• ISSNs: 1751-8628

Abstract

Abstract: 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. Abstract : After investigating the training process of neural network formed by unrolling a original LDPC code decoder, we proposed a novel method of generating the needed feeding data, and exposed the close connection between loss and decoding metrics of a code. Furthermore, it was verified in extensive simulation that the quantity of training parameter impacts decoder performance not so much as its placement.