Improved polar decoder based on deep learning

TLDR

Deep learning has recently shown strong competitiveness for improving polar code decoding, but conventional DNNs suffer from prohibitive training and computation complexity, limiting them to very short code lengths. This paper addresses deep learning decoding challenges by proposing a multiple‑scaled belief propagation algorithm and a low‑complexity neural network decoder applicable to any code length. The authors develop the multiple‑scaled BP algorithm and the neural network decoder, and implement a hardware architecture with folding techniques that cuts computation cost by roughly 50%. The proposed NND requires only a small set of zero codewords for training, achieves lower BER than conventional BP with fewer iterations, and its hardware implementation reduces cost by about 50%.

Abstract

Deep learning recently shows strong competitiveness to improve polar code decoding. However, suffering from prohibitive training and computation complexity, the conventional deep neural network (DNN) is only possible for very short code length. In this paper, the main problems of deep learning in decoding are well solved. We first present the multiple scaled belief propagation (BP) algorithm, aiming at obtaining faster convergence and better performance. Based on this, deep neural network decoder (NND) with low complexity and latency, is proposed for any code length. The training only requires a small set of zero codewords. Besides, its computation complexity is close to the original BP. Experiment results show that the proposed (64,32) NND with 5 iterations achieves even lower bit error rate (BER) than the 30-iteration conventional BP and (512, 256) NND also outperforms conventional BP decoder with same iterations. The hardware architecture of basic computation block is given and folding technique is also considered, saving about 50% hardware cost.