1

This possibly is a very stupid question, but i have not been able to find the answer on the internet and have got no clue which keywords to use while searching.

What's the meaning of $\mathbb{E}_{x \sim p_{data}(h)} [...]$ Where ... is some function.

in context like

$\mathbb{E}_{x \sim p_{data}}[...] + \mathbb{E}_{x \sim p_{z}} [...]$

and

$\mathbb{E}_{x \sim p_{data}(h)} [...] + \mathbb{E}_{x \sim p_{z}(z)} [...]$

This is written in a lot of papers, but what does it mean?

deKeijzer
  • 113
  • 5

1 Answers1

0

I'm supposing you are reading the paper on GAN where $[...]$ is the loss function. $\mathbb{E}_{x \sim p_{data}(h)} [...]$ now represents the expected loss where the data is taken from the distribution $p_{data}(h)$.

Edit:
$\mathbb{E}[ ]$ is an operator which gives the expected value of the function on which it is acted upon. When specifically acted on the loss function it is called risk or empirical loss. And L(z) that you are asking is the loss.

For more on its definition and explanation, see the below questions:

Risk function or expected loss function

why do we calculate risk when we already have loss functions?

bkshi
  • 2,235
  • 2
  • 11
  • 23
  • Yes that's correct. Is it the same meaning as $\mathcal{L(z)}$, used in some image inpainting with DCGAN papers? However, what is the formal mathematical meaning of the notation? I guess there must be a more formal definition then just 'this is the loss when $x$ is from $p_{data}(h)$ ? What's the meaning of $\mathbb{E}$ in this context? just 'loss' ? – deKeijzer Feb 05 '19 at 14:27
  • I thought you were just looking for the notation. Well, I have added more information as you asked. – bkshi Feb 06 '19 at 08:12