aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--posts/2019-02-14-raise-your-elbo.md6
1 files changed, 5 insertions, 1 deletions
diff --git a/posts/2019-02-14-raise-your-elbo.md b/posts/2019-02-14-raise-your-elbo.md
index d4e2227..2a59c53 100644
--- a/posts/2019-02-14-raise-your-elbo.md
+++ b/posts/2019-02-14-raise-your-elbo.md
@@ -60,7 +60,7 @@ It achieves minimum $0$ when $p = q$.
If $p$ can be further written as
-$$p(x) = {w(x) \over Z}$$
+$$p(x) = {w(x) \over Z}, \qquad (0)$$
where $Z$ is a normaliser, then
@@ -96,6 +96,8 @@ w &= p(x, z; \theta)\\
q &= q(z)
\end{aligned}$$
+and the dummy variable $x$ in (0) is substituted with $z$.
+
Fully Bayesian VI, often just called VI, has the following
instantiations:
@@ -105,6 +107,8 @@ w &= p(x, z, \theta) \\
q &= q(z, \theta)
\end{aligned}$$
+and $x$ in (0) is substituted with $(z, \theta)$.
+
In both cases $\theta$ are parameters and $z$ are latent variables.
**Remark on the naming of things**.