On the Belief Propagation and its Dynamics
Jean-Christophe Sibel, Sylvain Reynal · HAL (Le Centre pour la Communication Scientifique Directe) · 2012
In this study, we investigate the Belief Propagation (BP) algorithm proposed in [1] as an inference method to estimate the joint probability distribution of a given Bayesian network. Many works have raised up the use of BP in numerous applications as in image processing, computer vision, neural network, statistical physics and channel coding.(...) This study helps to classify four distinguished dynamical be- haviors of the BP along the SNR values. It confirms that the BP is widely infuenced by the topology of the considered LDPC code. The estimators, ei- ther well-known or new, provide qualitative and quantitative descriptions of the algorithm, especially for the complex dynamics.