aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorYuchen Pei <me@ypei.me>2019-02-19 19:09:45 +0100
committerYuchen Pei <me@ypei.me>2019-02-19 19:09:45 +0100
commit1e3b8940b5ff8914c09ea99cc411167206faf937 (patch)
tree2a456d5108b0c08e944c5d1c8da3b70f03f965b6
parent924140fa6851dfe630b390b2e60f52c6cb2ab34b (diff)
more minor edits
-rw-r--r--posts/2019-02-14-raise-your-elbo.md4
1 files changed, 2 insertions, 2 deletions
diff --git a/posts/2019-02-14-raise-your-elbo.md b/posts/2019-02-14-raise-your-elbo.md
index 64dfd81..cf3911f 100644
--- a/posts/2019-02-14-raise-your-elbo.md
+++ b/posts/2019-02-14-raise-your-elbo.md
@@ -189,7 +189,7 @@ A different way of describing EM, which will be useful in hidden Markov
model is:
- At E-step, one writes down the formula
- $$\sum_i \mathbb E_{p(z | x_{i}; \theta_t)} \log p(x_{i}, z; \theta). \qquad (2.5)$$
+ $$\sum_i \mathbb E_{p(z_i | x_{i}; \theta_t)} \log p(x_{i}, z_i; \theta). \qquad (2.5)$$
- At M-setp, one finds $\theta_{t + 1}$ to be the $\theta$ that
maximises the above formula.
@@ -208,7 +208,7 @@ so we write $\eta_k = (\mu_k, \Sigma_k)$,
During E-step, the $q(z_i)$ can be directly computed using Bayes'
theorem:
-$$r_{ik} = \mathbb P(z = k | x = x_{i}; \theta_t)
+$$r_{ik} = \mathbb P(z_i = k | x = x_{i}; \theta_t)
= {g_{\mu_{t, k}, \Sigma_{t, k}} (x_{i}) \pi_{t, k} \over \sum_{j = 1 : n_z} g_{\mu_{t, j}, \Sigma_{t, j}} (x_{i}) \pi_{t, j}},$$
where