Decoding Short LDPC Codes via BP-RNN Diversity and Reliability-Based Post-Processing

This paper investigates decoder diversity architectures for short low-density parity-check (LDPC) codes, based on recurrent neural network (RNN) models of the belief-propagation (BP) algorithm. We propose a new approach to achieve in waterfall region, by specializing BP-RNN decoders specific classes errors, with absorbing set support. further combine our an ordered statistics decoding (OSD) post-processing step, which effectively leverages bit-error rate optimization deriving from use binary cross-entropy loss function. show that single specialized combines better than BP OSD step. Moreover, combining brought multiple decoders, provides efficient way bridge gap maximum likelihood decoding.