The most significant reason for this is the computational complexity of learning the structure of the networks, a problem that has been shown to be nondeterministic polynomial time (NP)–hard (Chickering 1996), implying that an exact computational solution is not possible.
