IEEE Communications Letters
Parallel Weighted Gradient-Descent-Bit-Flipping Decoding With Momentum
Xinran Sun, Xiaofu Wu
Abstract
Gradient descent bit-flipping decoding with momentum (GDBF-w/M) has demonstrated excellent performance for LDPC codes, yet it often requires a large number of decoding iterations to converge. This letter proposes a novel parallel weighted gradient-descent-bit-flipping decoding with momentum (PW-GDBF-w/M) to improve both the convergence rate and the final decoding performance. Simulation results show that the proposed PW-GDBF-w/M significantly outperforms the exsting GDBF-w/M algorithms in terms of both performance and convergence, and achieves results very close to those of belief-propagation (BP) decoding for the (2048, 1723) 802.3an LDPC code.