Updating the Model
n
Given a new data instance (
Q
:=
q, x
):
n
Update P(G,
q
G
) to get posterior P(G,
q
G
|
Q
:=
q, x
)
n
P(G,
q
G
|
Q
:=
q, x
) is our new model
n
A distribution over graphs and parameters
n
Given a model P(G,
q
G
) in principle we can obtain P(X
®
Y)