Directed graphical models, popularly known as Bayesian networks, are an important family of probabilistic graphical models. They are a convenience method to express complicated relationships among random variables. Being graphical, Bayesian networks offer a simple way to visualize probabilistic dependencies between variables. Moreover, complex computations such as those required to perform learning and inference can be expressed in terms of graphical operations.
The undirected counterparts to Bayesian networks are known as Markov random fields, or simply undirected graphical models. We cover those separately.