资 源 简 介
应用背景We provide some example Matlab code as a supplement to the paper [6].
This technical report is not intended as a standalone introduction to the
belief propagation algorithm, but instead only aims to provide some technical
material, which didn’t fit into the paper.关键技术In the graphical output generated in the previous section we observe that some trajectories
seem to cease to exist and then possibly reappear later on. This effect is caused by
round-off errors due to the numerical precision of the computations in Matlab. For very
large arguments the tanh-function returns 1, and atanh(1) = 1. In theory though, finite
arguments always give finite outputs, but number representation limits Matlab in actually
producing accurate results after sometimes only a few iterations. Countermeasures
employed, e.g., in [3], include to apply some saturation function after each iteration to
