
110 Bayesian Networks: With Examples in R
Algorithm 4.3 Sparse Candidate Algorithm
1. Choose a network structure G over V, usually (but not necessarily)
empty.
2. Repeat the following steps until convergence:
(a) restrict: select a set C
i
of candidate parents for each node
X
i
∈ V, which must include the parents of X
i
in G;
(b) maximise: find the network structure G
∗
that maximises
Score(G
∗
) among the networks in which the parents of each
node X
i
are included in the corresponding set C
i
;
(c) set G = G
∗
.
3. Return the DAG G.
tures in a wide variety of situations. The two b est-known members of this
family are the Sparse Candidate algorithm (SC) by Friedman et