作者: Gérard Battail , Rémi Sfez
关键词: Linear code 、 Algorithm 、 Pattern recognition 、 Mathematics 、 Inverse-chi-squared distribution 、 Principle of maximum entropy 、 Artificial intelligence 、 Decoding methods 、 Parity bit 、 Maximum entropy probability distribution 、 Kullback–Leibler divergence 、 Posterior probability
摘要: Several successive decodings of cascaded codes become possible in principle without information loss if the decoding task is extended to determine a posterior probability distribution on codewords. Kullback cross-entropy minimization considered as means implementing it. Its practical use, however, demands some kind simplification. We propose look for separable form with respect symbols, which leads output same its input. As an illustration these ideas, we iterated product parity-check results vanishingly small error provided channel signal-to-noise ratio larger than threshold. Interpreting single linear code parity checks, ideas lead simple and efficient algorithm.