Timeline for Why is the log probability replaced with the importance sampling in the loss function?
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 7, 2021 at 1:25 | comment | added | BreakDS | Yep, so basically the $log$ is gone because importance sampling cancels it out with the gradient of $log$. | |
| Jul 18, 2020 at 15:08 | history | edited | nbro | CC BY-SA 4.0 |
images to latex
|
| Apr 12, 2020 at 0:26 | comment | added | Costa Huang | Nvm, I think your solution checks out. | |
| Apr 11, 2020 at 20:44 | comment | added | Costa Huang | Hi, nice intuition, but I think there is an error with your derivation on the 4th step in the second equation. When you take the $\nabla_{\theta} \pi_{\theta}(a_t|s_t)$, why isn't the derivative taken for all the other $\pi_{\theta}(a_t|s_t)$? | |
| Jul 5, 2019 at 13:35 | review | Late answers | |||
| Jul 12, 2019 at 13:20 | |||||
| Jul 5, 2019 at 13:20 | review | First posts | |||
| Jul 5, 2019 at 21:28 | |||||
| Jul 5, 2019 at 13:16 | history | answered | mglss | CC BY-SA 4.0 |