Bayesian networks

Question 1

Why do we want to use Bayesian Networks?

Answer 1

To compactly represent probability distributions (belief /likelihood of observations)

Question 2

Does a fully-connected Bayesian network provide any advantage over a full joint probability table representation?

Answer 2

No, they would be the same.

Question 3

What is the Markov blanket of a Bayesian network node “n”?

Answer 3

It is the set of nodes that makes “n” conditionally independent of all the others. This set of nodes is made of: “n”s parents, children and children’s parents.