Figure 1.

Bayesian networks with variables X and E denoting past data, and Y and F denoting future data. Both past and future data depend on and are related through, in (i), parameter θ, belonging to the parameter space Θ, and in (ii), a variable H denoting discrete hypotheses. The dashed edges denote the process when going from past data to the parameter, called inference, whereas the dotted edge illustrates the process of going to future observations.