Generalized Belief Propagation Decoding of Polar Codes
Nikolai Iakuba, Peter Trifonov · 2023
A generalized belief propagation algorithm is proposed for decoding of polar codes. The proposed approach relies on offline successive transformations of the factor graph, which merge parity check nodes into check nodes corresponding to some non-trivial codes. It was shown, that the constructed factor graph for (2048, 1024) polar code provides 0.25 performance gain under generalized belief propagation decoding with 5 iterations and lower worst case complexity compared to state-of-the-art belief propagation decoders of polar codes.