Probabilistic reasoning in intelligent systems: Networks of plausible inference