Multi-connection BP Decoding for Polar Codes
Liuchang Yang, Dianhong Wang, Zhongxiu Feng, Yangyang Liu, Lixia Xiao · 2022
In this paper, we develop a multi-connection belief propagation (BP) decoding algorithm for polar codes, which employs the idea of the residual neural network to accelerate the convergence. Specifically, multi-connection BP decoding builds on the standard BP decoding by fusing the soft information of the current decoding iteration and the past decoding iterations proportionally according to the damping factor during each iteration. Moreover, we adopt the particle swarm optimization algorithm to obtain the optimal value of the damping factor to balance a trade-off between the error rate performance and decoding complexity. The suggested approach can outperform the standard BP decoding with lower iterations, according to simulation findings, which demonstrate that it can reduce error rates.