Bayesian point estimation II
n
We do not know the true
q
n
Density p
represents our optimal beliefs over
q
n
Choose
q
that minimizes the expected loss:
n
n
q
= argmin
q
n
n
~
n
Call
q
the
Bayesian point estimate
n
We will use
KL divergence
as our loss
n
q
=E(
q
)
~
~