I've never seen the following notation before:
What does it mean?
Example in the cost function here
Asked
Active
Viewed 56 times
0
-
from the deep learning book? which page? – Haitao Du Sep 20 '17 at 18:54
-
1Notation is just a convention -- this will need more context to decipher. – Sycorax Sep 20 '17 at 19:10
-
2One guess is that $p_{\text{data}}$ means the empirical distribution. It hardly matters, because the expectation of the constant $-\log 2$ is $-\log 2$, the distribution do not matter! – kjetil b halvorsen Sep 20 '17 at 19:30
-
Looks like it means negative log two. – Matthew Drury Sep 20 '17 at 23:15
-
Updated with an example – Dean Sep 21 '17 at 07:27
-
In the description of the EM algorithm on Wikipedia, people use to do the same stupid things: https://en.wikipedia.org/wiki/Expectation%E2%80%93maximization_algorithm. Here they write $E_{Z|X}[f(Z,X)]$ and they actually mean is the (factorization of the) conditional expectation $E[f(Z,X)|X=x]$ and I assume that they want to express something similar here... maybe it is $E[\text{log}p_{\theta}(X|Z) | X=x]$ or maybe restricted to $Z=z$? – Fabian Werner Sep 21 '17 at 07:41