For this example, we are going to use hmmlearn, which is a package for HMM computations (see the information box at the end of this section for further details). For simplicity, let's consider the airport example discussed in the paragraph about the Bayesian networks, and let's suppose we have a single hidden variable that represents the weather (of course, this is not a real hidden variable!), modeled as a multinomial distribution with two components (good and rough).
We observe the arrival time of our flight London-Rome (which partially depends on the weather conditions), and we want to train an HMM to infer future states and compute the posterior probability of hidden states corresponding to a given ...