Belief propagation vs. TAP for decoding corrupted messagesY. Kabashima and D. SaadEurophys. Lett., 44 5 (1998) 668-674DOI: https://doi.org/10.1209/epl/i1998-00524-7
