AlxndrMlk/pgms

Notes on probabilistic graphical models (PGMs)

MIT

PGMs

Notes on probabilistic graphical models (PGMs)

Bayesian networks

A Bayesian network (BN):

is a directed DAG $G$
nodes represent random variables $X_1, ..., X_n$
for each node we have a conditional probablility distribution (CPD) represented by $P(X_i|Parents_G(X_i))$
a Bayesian network represents a joint distribution of all its nodes via the chain rule for BNs: $P(X_i, ..., X_n) = \prod_i^N P(X_i|Parents_G(X_i))$