Guessing Random Additive Noise Decoding with Quantized Soft Information
Peihong Yuan, Ken R. Duffy, Evan P. Gabhart, Muriel Médard · 2023
In this work, we introduce discretized soft GRAND (DS-GRAND) based on dynamic programming (DP), which utilizes quantized soft information. Typical quantization values for per-information bit soft information in decoding chips range from 3 to 5 bits. Our simulations indicate that DSGRAND performs within 0.25 dB and 0.1 dB of maximum-likelihood (ML) decoding with 2 and 3 bit soft information quantizers, respectively. We analyze the memory requirements and computational complexity of DSGRAND, demonstrating that for the CA-SCL Polar decoder with a list size of 128, which closely approaches DSGRAND performance, DSGRAND outperforms CA-SCL by an order of magnitude in time complexity and two orders of magnitude in space complexity.