I am looking at the machine learning paper, and I came across expected-value in the following notation:

$E_{s_t \sim \rho^B, \alpha \sim \beta, r_t \sim E}[(Q(s_t, a_t |\theta^Q)-y_t)^2]$

(eqn 4) on https://arxiv.org/pdf/1509.02971.pdf

My Question is, Does it mean that the expectation is taken over $s_t, \alpha, $ and $ r_t$? what is that $_{s_t \sim \rho^B, \alpha \sim \beta, r_t \sim E}$ means? Does it mean that the expectation is taken over $s_t, \alpha, $ and $ r_t$?



Yes; it means that you take the expectation over those variables. The ~ sign is to show you what "space" they live in; so the variable $s_t$ comes from the space $\rho^\beta$. So, when you take an expectation over those, they will disappear, and the remaining quantity will be a function of theta. (note that $y_t$ is also a function of $\theta$ as they mention it in their paper; the other parameters it depends on also disappear after taking expectation)

  • $\begingroup$ how do you take expectation over multiple variables $\endgroup$ – kou Apr 3 '18 at 6:29
  • $\begingroup$ Have you taken a class called multivariable calculus? $\endgroup$ – E-A Apr 3 '18 at 6:32
  • 1
    $\begingroup$ I tried to edit "perimeters" to say "parameters", but Stack wouldn't let me change less than 6 characters. $\endgroup$ – Joe May 19 '20 at 15:36
  • $\begingroup$ Yup; thanks! Appreciated. $\endgroup$ – E-A May 19 '20 at 23:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.